Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 193 appraisals
Journal of medical Internet research
Magnetic Resonance Imaging-Based Artificial Intelligence in Predicting Prostate Cancer Biochemical Recurrence: Systematic Review and Meta-Analysis
BACKGROUND: Artificial intelligence (AI) has emerged as a promising tool for prostate cancer (PCa) risk stratification and outcome prediction. However, current studies often lack multicenter external validation, have limited sample sizes, present significant intermodel variability, and face overfitting concerns. OBJECTIVE: This study aimed to comprehensively evaluate the diagnostic performance of magnetic resonance imaging (MRI)-based AI models in predicting biochemical recurrence (BCR) of PCa. METHODS: Systematic searches were conducted in the PubMed, Embase, Web of Science, and Cochrane Library databases up to January 13, 2026. Studies were included that involved participants diagnosed with PCa, used MRI-based AI for predicting BCR, and had clearly defined reference standards. The quality of the included studies was assessed using the PROBAST+AI tool. A bivariate random effects model was used to pool sensitivity, specificity, and area under the curve (AUC) statistics. RESULTS: A total of 28 studies were included, with 2623 patients in internal validation and 1134 patients in external validation. Diagnostic contingency tables were reconstructed from published performance metrics for most studies, while others were extracted from receiver operating characteristic curves due to the lack of direct reporting. In the internal validation set, pooled sensitivity was 0.80 (95% CI 0.73-0.86; prediction interval [PI] 0.48-0.99), specificity was 0.83 (95% CI 0.77-0.89; PI 0.49-1.00), and AUC was 0.86 (95% CI 0.83-0.89; PI 0.74-0.99). In the external validation set, pooled sensitivity was 0.82 (95% CI 0.72-0.91; PI 0.54-0.99), specificity was 0.83 (95% CI 0.71-0.92; PI 0.49-1.00), and AUC was 0.84 (95% CI 0.79-0.90; PI 0.70-0.98). No statistically significant differences were observed between internal and external validation in sensitivity (P=.73), specificity (P>.99), AUC (P=.53), or diagnostic odds ratio (P=.98). Medical Net and Extreme Gradient Boosting achieved the highest sensitivity and AUC, whereas multiple kernel learning and support vector machine had the highest specificity. Subgroup and meta-regression analyses suggested that AI method, model type, timing of MRI acquisition, and treatment modality may contribute to heterogeneity. CONCLUSIONS: This meta-analysis innovatively realizes the quantitative direct comparison of MRI-based AI model performance between internal and external validation cohorts for PCa BCR prediction. It comprehensively evaluates AI performance across diverse PCa treatment modalities and integrates machine learning and deep learning approaches. For the field, it identifies key sources of performance heterogeneity (eg, MRI acquisition timing and treatment modality) and quantifies the sensitivity-specificity trade-off in integrated radiomic-clinical models, advancing the systematic understanding of MRI-based AI for BCR prediction. In real-world practice, it provides actionable guidance to prioritize pretreatment MRI for AI model development and clinical BCR assessment and underscores the urgent need for standardized imaging protocols and prospective multicenter studies, laying a foundation for the safe clinical translation of these AI tools as adjunctive decision support instruments.
8 July 2026
Read appraisal →Korean journal of radiology
Agentic Artificial Intelligence for the Automated Generation of Accurate Summary Podcasts of Radiology Research Papers
OBJECTIVE: To evaluate whether a custom agentic artificial intelligence (AI) pipeline can overcome the limitations of general-purpose large language model tools, when compared with a generic commercial tool (Google NotebookLM [NBLM]), for generating podcast-style summaries of radiology research articles. MATERIALS AND METHODS: Twenty-two PDF-format original research articles published in the April 2025 issue of Radiology were processed using our Programmable, Phoneme-Aware PDF-to-Podcast Pipeline (P5) and NBLM to generate 44 audio episodes. P5 utilizes a multi-agent workflow for script generation, quality assurance, pronunciation enhancement, and audio synthesis. Four radiologists from a pool of 25 (7 generalists and 18 specialists) were randomly assigned to evaluate each blinded audio episode, yielding 176 total evaluations. The primary outcomes were the number of hallucinations (factual errors) per episode and the percentage of hallucination-free episodes. Secondary outcomes included the number of inappropriate statements, mispronunciations, and flow disruptions; the composite quality score (Quality Assessment of Educational Podcasts [QAEP]); the key results coverage score; and overall listener preference. Data were analyzed using generalized linear mixed models. RESULTS: The P5 method produced significantly fewer hallucinations per episode compared with NBLM (mean, 0.32 vs. 0.93; P < 0.001) and a higher proportion of hallucination-free episodes (71.6% [63/88] vs. 56.8% [50/88]; P = 0.013), consistently across generalists and specialists. P5 demonstrated significantly fewer mispronunciations (mean, 0.11 vs. 1.62; P < 0.001) and flow disruptions (mean, 0.19 vs. 1.06; P < 0.001) per episode, as well as higher mean QAEP composite scores (4.62 vs. 4.28; P < 0.001), compared with NBLM, consistently across generalists and specialists. Raters preferred P5 over NBLM in 72.7% (64/88) of comparisons (P = 0.003). CONCLUSION: Our custom agentic AI pipeline generated podcast-style summaries of radiology research articles with significantly higher quality and greater listener preference than the generic commercial tool.
3 July 2026
Read appraisal →Medicine
Artificial intelligence-guided clavicle fracture detection on plain radiographs: A retrospective diagnostic accuracy study.
Extensive research exists on artificial intelligence (AI)-assisted fracture detection in various anatomical regions, but clavicle fractures remain comparatively understudied. This study aimed to compare the diagnostic performance of AI-guided clavicle fracture detection on plain radiographs with that of an experienced orthopedic surgeon. We retrospectively analyzed all clavicle radiographs obtained at our institution between September 11,2023 and October 31,2024 following the clinical implementation of AI-guided fracture detection. Radiological findings confirmed by a senior radiologist served as the reference standard. AI performance was additionally compared with the assessments of an experienced, blinded orthopedic surgeon. Diagnostic performance metrics included accuracy, sensitivity, specificity, Cohen kappa, F1 score, and Youden index. A total of 367 clavicle radiographs were included, comprising 186 anteroposterior (AP) and 181 second-plane views. The mean patient age was 44.6 ± 26.2 years, and 59.4% were male. According to the reference standard, clavicle fractures were present in 96 AP radiographs (51.6%) and 95 second-plane radiographs (52.5%). On AP radiographs, the AI system achieved an accuracy of 97.31%, sensitivity of 95.83%, specificity of 98.89%, Cohen kappa of 0.95, and F1 score of 0.97. Diagnostic performance on second-plane radiographs was similarly high and overall comparable to that of the experienced orthopedic surgeon. This retrospective diagnostic accuracy study shows that AI-guided clavicle fracture detection is highly accurate and reliable, with performance approaching that of an experienced orthopedic surgeon. These findings support the potential role of AI as a diagnostic support tool in clavicle fracture detection, while prospective and multicenter validation remains warranted.
28 June 2026
Read appraisal →European radiology experimental
Digital twin technologies in prostate cancer as a frontier for precision medicine
Prostate cancer (PCa) is the most frequently diagnosed malignancy among men and presents major clinical and socioeconomic challenges worldwide. Despite advances in early detection, imaging, and therapy, managing PCa remains complex due to disease heterogeneity, risks of overdiagnosis, and overtreatment. Digital twin (DT) technologies might represent an emerging conceptual framework aimed at supporting dynamic, patient-specific virtual modeling for personalized clinical decision-making. By integrating multimodal clinical, imaging, molecular, and physiological data, DTs can simulate disease progression, predict treatment responses, and support proactive, adaptive care. This perspective explores the conceptual framework for DT ecosystems in PCa, highlighting potential clinical impacts, infrastructural requirements, and barriers to implementation. Harnessing DTs could impact PCa management into a truly predictive, personalized, and participatory approach, improving outcomes and optimizing healthcare resource utilization globally. RELEVANCE STATEMENT: DT technologies may enable personalized, predictive PCa management by integrating multimodal patient data to guide diagnosis, treatment selection, and monitoring, with the potential to improve outcomes, reduce overtreatment, and optimize healthcare resource utilization KEY POINTS: DTs create virtual patient models to personalize PCa care. They integrate imaging, clinical, and molecular data into one system. This approach may reduce overdiagnosis and unnecessary treatments.
20 June 2026
Read appraisal →Journal of neural engineering
Uncertainty-aware multi-path framework with dynamic arbitration for Alzheimer's disease MRI classification
Objective.As a progressive neurodegenerative disorder, Alzheimer's disease (AD) requires early and accurate diagnosis to delay pathological progression and improve clinical outcomes. Recently, Deep-learning methods based on structural MRI (sMRI) have been widely used for computer-aided AD diagnosis. However, reliable deployment remains challenging due to its sample-specific heterogeneity and the lack of explicit mechanisms for handling prediction uncertainty. In addition, common test-time perturbations and augmentation-based inference strategies could introduce anatomical inconsistency in medical images, leading to unstable decisions for hard cases.Approach.Accordingly, TriPathNet with Arbiter, as an uncertainty-aware multi-path framework, is proposed in the study. The framework learns spatially complementary representations via multi-view parallel encoding with metric-guided coordination. In addition, an augmentation-aware arbitration mechanism is introduced to adaptively correct predictions for high-uncertainty samples during inference.Main results.On the ADNI dataset for the AD vs. cognitively normal task, the proposed method achieves 98.85% accuracy, 99.18% sensitivity, 97.06% specificity, and 99.61% AUC, yielding satisfactory performance.Significance.The results indicate that TriPathNet with Arbiter provides an accurate and robust solution for sMRI-based computer-aided AD diagnosis, maintaining stable performance under test-time perturbations.
20 June 2026
Read appraisal →Spine
Differentiating Spinal Deformity Versus Leg Length Discrepancy in Patients With Pelvic Obliquity : Systematic Review of Diagnostic Methods, Imaging, and Clinical Management
STUDY DESIGN: Systematic review, PROSPERO ID: (CRD42023112392). OBJECTIVE: We aim to provide physicians with an evidence-based stepwise approach to distinguishing the primary cause of plane pelvic obliquity (PO) to guide treatment. SUMMARY OF BACKGROUND DATA: PO is a frequent clinical finding in both spinal deformity (SD) and leg length discrepancy (LLD) but distinguishing between these etiologies remains challenging due to overlapping clinical and radiographic features and potential multifactorial origins. Misdiagnosis may delay care and worsen patient outcomes. MATERIALS AND METHODS: PubMed, Embase, Scopus, and Web of Science were searched from inception to July 2025 for English-language studies evaluating coronal PO in the context of LLD and/or SD that explicitly compared or differentiated limb-origin from spine-origin obliquity using radiographic, advanced imaging, or validated clinical assessments and reported diagnostic accuracy, discriminative features, or management impact. Two reviewers independently screened studies, extracted data, and performed ROBINS-I risk assessment and Newcastle-Ottawa Scale quality assessment. Owing to marked heterogeneity in populations, measurement protocols, and thresholds, findings were synthesized qualitatively with stratification by age group where possible. RESULTS: Of 436 records, 24 cohort studies met the inclusion criteria. Standing anteroposterior (AP) radiographs reliably quantified both LLD and PO, while advanced imaging provided high reproducibility for subtle or ambiguous cases. Radiographic block or shoe lift correction tests distinguished limb-origin and spine-origin PO through immediate postural correction. Key discriminators included a lumbar curve or PO resolving with block/lift correction suggesting LLD, versus persistent deformity despite correction suggesting spinal etiology. Difficulty remains in diagnosing and treating mixed etiology patients. CONCLUSION: Accurate differentiation of spinal versus limb-driven PO begins with a structured diagnostic approach. AP radiographs, followed by functional block tests, and advanced imaging, may increase diagnostic confidence and decrease unnecessary interventions. Particularly in patients with both LLD and coronal plane SD, prospective studies are required to standardize measurement protocols and improve long-term outcomes. LEVEL OF EVIDENCE: Level IV.
16 June 2026
Read appraisal →Journal of neurology, neurosurgery, and psychiatry
Clinically reported covert cerebrovascular disease and risk of neurological disease: a whole-population cohort of 367 988 people using natural language processing
BACKGROUND: The relevance of covert cerebrovascular disease (CCD) in practice is uncertain, partly because estimation of risk in whole clinical populations is difficult. Studies have had success extracting CCD from clinical text using natural language processing (NLP), though they have been limited to specific CCD phenotypes. Here, we used NLP to measure multiple clinically-reported CCD phenotypes in a large clinical cohort and estimated subsequent disease risk in health record data. METHODS: From all people with brain imaging in Scotland (2010-2018), we selected people with no prior hospitalisation for neurological disease (n=367 988). NLP of imaging reports identified: white matter hypoattenuation or hyperintensities (WMH), lacunes, cortical infarcts and cerebral atrophy. Adjusted HRs (aHRs) were estimated between each phenotype and stroke, dementia and Parkinson's disease (conditions previously associated with CCD), epilepsy and colorectal cancer (control conditions). RESULTS: For each phenotype, the aHR of stroke was WMH 1.4 (95% CI 1.3-1.4), lacunes 1.6 (1.5-1.6), cortical infarct 1.8 (1.7-1.9) and cerebral atrophy 1.1 (1.0-1.1). The aHR of dementia was WMH 1.3 (1.3-1.3), lacunes 1.0 (0.9-1.0), cortical infarct 1.1 (1.1-1.2) and cerebral atrophy 1.7 (1.7-1.8). The aHR of Parkinson's disease was WMH 1.1 (1.0-1.2), lacunes 1.1 (0.9-1.2), cortical infarct 0.7 (0.6-0.9) and cerebral atrophy 1.4 (1.3-1.5). The aHRs between CCD phenotypes and epilepsy and colorectal cancer were around the null. CONCLUSION: CCD and atrophy have implications for future disease risk and can be identified at scale using NLP of clinical reports. Prevention of neurological disease in people with CCD should be a priority for healthcare policy makers.
14 June 2026
Read appraisal →JMIR formative research
Automated Machine Learning Frameworks for Radiomics: Comparative Evaluation Study
BACKGROUND: Automated machine learning (AutoML) frameworks can lower technical barriers for predictive and prognostic model development in radiomics by enabling researchers without programming expertise to build models. However, their effectiveness in addressing radiomics-specific challenges remains unclear. OBJECTIVE: This study aimed to evaluate the performance, efficiency, and accessibility of general-purpose and radiomics-specific AutoML frameworks on diverse radiomics classification tasks, thereby guiding researchers and highlighting development needs for radiomics. METHODS: A total of 10 public and private radiomics datasets with varied imaging modalities (computed tomography and magnetic resonance imaging), sizes, anatomies, and end points were used. Six general-purpose and 5 radiomics-specific frameworks were tested with predefined parameters using standardized cross-validation. Evaluation metrics included area under the receiver operating characteristic curve, runtime, and qualitative aspects related to software status, accessibility, and interpretability. RESULTS: Simplatab, a radiomics-specific tool with a no-code interface, achieved the best overall balance between performance and computational efficiency, recording the highest average test area under the receiver operating characteristic curve (mean 78.46%, SD 12.22%) with a moderate runtime (1.1 h). However, its performance was not statistically superior to the most intensive general-purpose solutions. Most radiomics-specific frameworks were excluded from the performance analysis due to obsolescence, extensive programming requirements, or computational inefficiency. Conversely, general-purpose frameworks demonstrated higher accessibility and ease of implementation. CONCLUSIONS: While no single framework demonstrated absolute predictive superiority, Simplatab provides an effective balance of performance, efficiency, and accessibility for radiomics classification problems. However, continued efforts are needed to further mature AutoML solutions in the radiomics domain.
14 June 2026
Read appraisal →Scientific reports
Performance of GPT-based large language models in hepatocellular carcinoma stratification: liver function assessment, BCLC staging, and treatment recommendations.
Large language models (LLMs) like GPT have been proposed to support complex clinical decision-making. This study evaluated the performance of GPT-based LLM in analyzing clinical, radiological, and laboratory data from patients with hepatocellular carcinoma (HCC) to assess liver function, assign BCLC stage, and recommend treatment. Data from 106 HCC patients (82% male, median age 65 [22-86]) were compiled into anonymized integrated reports. Four GPT-versions (4, o1, o3, 5.4) were prompted-using both short and long instructions-to calculate MELD, ALBI, and Child-Pugh scores, assign BCLC stage, and generate treatment recommendations based on current guidelines. Outputs were compared to expert consensus and tumor board decisions. Errors were categorized by type and source. Time and cost analyses compared GPT to clinical staff. All GPT versions achieved high accuracy (> 85%) in liver function assessment, with MELD calculation being the most error-prone. BCLC staging accuracy ranged from 46.2% (version 4) to 84.0% (o3), with misclassification of radiological reports as the main error source. Reasoning-optimized models (o1, o3) performed best for treatment recommendations, achieving an overall accuracy (correct suggestions and acceptable alternatives) of up to 90.6%. In 9-14% of cases, GPT suggestions were retrospectively more guideline-concordant than tumor board decisions. GPT processing was significantly faster and reduced costs by approximately 300- to 1300-fold compared to clinical staff. GPT-based LLMs show potential as decision-support tools for liver function assessment, BCLC staging, and treatment guidance in HCC. Particularly with reasoning-optimized models and detailed prompting, LLMs may serve as valuable adjuncts in multidisciplinary HCC workflows. However, a non-negligible error rate requires expert oversight and further model refinement.
14 June 2026
Read appraisal →Biomedical physics & engineering express
Enhancing brain tumor classification with a simplified CNN through hyperparameter optimization
Brain tumors remain a major public health challenge because of their high mortality rate and the need for timely and accurate diagnosis. Magnetic resonance imaging (MRI) is widely used for brain tumor assessment, yet conventional diagnostic workflows still depend largely on the subjective interpretation of radiologists, which may lead to inter-observer variability. In this context, deep learning approaches-particularly convolutional neural networks (CNNs)-have gained increasing attention for automated medical image analysis, as they are capable of learning discriminative features directly from imaging data and improving classification accuracy. Nevertheless, the performance of CNN-based models is strongly influenced by the choice of hyperparameters, whose optimization remains a non-trivial task due to complex and non-linear interactions within the network.This study investigates the impact of different hyperparameter optimization strategies on the performance of a simplified CNN model for brain tumor classification from MRI images. Three widely used optimization methods, Genetic Algorithms (GA), Bayesian Optimization (BO), and HyperBand-are considered and systematically compared. To emphasize efficiency and practical applicability, the CNN architecture is deliberately kept lightweight, aiming to reduce computational complexity while preserving sufficient modeling capacity. Experiments are carried out on a publicly available brain MRI dataset that includes glioma, meningioma, pituitary tumor, and healthy subjects.To account for performance variability and ensure statistical reliability, each optimized model is evaluated over ten independent runs using different random initializations. The reported results therefore include both the mean performance values and the corresponding standard deviations. The experimental analysis shows that the GA-based approach achieves a mean accuracy of 96.41%±0.36%, while BO improves this performance to 97.37%±0.30%. Among the evaluated strategies, HyperBand yields the highest average accuracy of 97.73%±0.48%, indicating its ability to consistently identify effective hyperparameter configurations for the proposed CNN architecture.
13 June 2026
Read appraisal →Placenta
Recent advances in multimodal approaches for investigating placental development and related diseases
The placenta is a vital organ that supports fetal growth and pregnancy maintenance. Its dysfunction is associated with severe complications, including preeclampsia, fetal growth restriction, and placenta accreta spectrum disorders. Traditional approaches to placental research have provided valuable insights but are limited in capturing the complexity of this dynamic organ. In recent years, multimodal approaches-integrating advanced imaging, single-cell and spatial omics, and artificial intelligence (AI)-have enabled comprehensive analyses of placental development and disease. These strategies offer improved diagnostic accuracy, deeper molecular understanding, and real-time assessment of placental function. This review summarizes recent advances in multimodal placental research, highlighting key technologies such as ultrasound and MRI, single-cell transcriptomics, spatial profiling, and AI-based prediction models. Particular emphasis is placed on contributions from Chinese research teams, who have developed novel platforms, atlases, and clinically relevant tools. We also discuss ongoing challenges, including data standardization, interpretability of AI models, and ethical considerations. Looking ahead, the integration of multimodal data with AI and wearable technologies holds promise for precision obstetrics and individualized pregnancy care. Together, these innovations are advancing both scientific understanding and clinical management of placenta-related disorders.
12 June 2026
Read appraisal →Nature communications
Adaptive Riemannian optimization for multi-scale diffeomorphic matching
Image matching is a fundamental task in quantitative biomedical and biological image analyses, enabling researchers to compare, integrate, and interpret imaging data across subjects, time points, modalities, and experimental conditions. Existing state-of-the-art registration methods are slow due to inefficient implementations and poor convergence rates because of the ill-conditioned nature of the optimization problem. Deep learning methods offer fast inference but require extensive training time, substantial inference memory, and fail to generalize across long-tailed distributions or diverse image modalities, necessitating costly retraining. We address these challenges by proposing FireANTs, a training-free, GPU-accelerated, multi-scale adaptive Riemannian optimization algorithm for fast and accurate dense diffeomorphic image matching. FireANTs more than doubles the speed of the community standard ANTs registration tool on a CPU, and is two orders of magnitude faster on a GPU. On the GPU, FireANTs performs competitively with deep learning methods on inference runtime while consuming up to 10 × less memory. FireANTs demonstrates robustness on a wide variety of matching problems across modalities, species, and organs, without any domain-specific training or tuning. Our framework allows hyperparameter grid search studies with less resources and time compared to traditional and deep learning registration algorithms alike.
12 June 2026
Read appraisal →Physics in medicine and biology
Unsupervised multimodal deformable medical image registration based on feature perceptual contrast learning
Objective.Multimodal medical image registration has extensive applications in clinical diagnosis and is fundamental for a series of medical analysis tasks. However, the presence of modality differences makes the registration process challenging. Existing methods often employ modality-independent feature descriptors that are sensitive to noise, or attempt to bridge differences within networks, which typically results in translation inaccuracies and misaligned anatomical information.Approach.In this paper, we propose a novel unsupervised approach utilizing a Feature Perceptual Contrast Learning Network (FP-net) to learn descriptors that bridge modality differences while accurately capturing common details. We unify the feature representation of anatomical information under homogeneous and heterogeneous intensity distributions through local sampling-based feature perceptual contrast learning and image reconstruction learning. The trained FP-net is subsequently employed to drive an unsupervised registration framework without requiring ground-truth deformation fields.Main results.We extensively evaluated our method on two public benchmarks: the BraTS 2021 dataset for brain T2-T1 and T1-T1ce registration, and the Learn2Reg 2021 dataset for challenging abdominal CT-MR registration. By passing multimodal image pairs with shape differences through the fixed FP-net, we generate optimization gradients that successfully update the registration network. Quantitative evaluations demonstrate our method's superiority over state-of-the-art baselines. Specifically, our model achieved a dice similarity coefficient (DSC) of 76.3% and 77.7% in tumor-bearing T2-T1 and T1-T1ce tasks, respectively. Furthermore, in the complex abdominal CT-MR task, it reached a DSC of 50.1%, significantly improving structural alignment.Significance.Our method effectively shifts the burden of bridging the modality gap away from the registration network, enabling standard U-Net architectures to achieve state-of-the-art deformable registration. This provides a robust, accurate, and easily deployable unsupervised solution for complex clinical multimodal image analysis.
12 June 2026
Read appraisal →Neurology
Hematopoietic Stem Cell Transplant and Brain Volume Changes in Adults With Sickle Cell Disease
BACKGROUND AND OBJECTIVES: Adults with sickle cell disease (SCD) are at risk of decline in brain health and cognition, even without clinical stroke. SCD-related changes on MRI include brain atrophy and T2 fluid attenuated inversion recovery (FLAIR) white matter hyperintensity (WMH) growth. Hematopoietic stem cell transplant (HSCT) can be curative, but conditioning regimens risk new complications, particularly for adults with accumulated organ damage. We tested whether successful HSCT for adults with SCD was associated with reduced brain atrophy or WMH growth. METHODS: In this cohort study, adults with severe SCD underwent HSCT as part of 5 research protocols at the National Institutes of Health Clinical Center, including autologous gene therapy, HLA-matched donor HSCs, and haploidentical HSCs. Serial brain MRI scans acquired before and after transplant for up to 12 years included 3D T1-weighted and axial FLAIR sequences. Longitudinal brain volumes, segmented using FreeSurfer and corrected for total intracranial volume (TICV), were used to measure brain parenchymal fraction (BPF). A deep learning algorithm segmented WMH volumes from FLAIR sequences. Volumes were log2-transformed as a percentage of TICV. We tested whether HSCT success (persistent engraftment vs graft failure) or change in hemoglobin (Hb) (post-pre-HSCT) was associated with the rate of brain atrophy or WMH growth using mixed-effects linear regression. Random effects included scanner, participant intercept, and slope, while fixed effects included age and sex. RESULTS: Seventy-seven participants (mean age 32 ± 9 years, 43% female) who underwent HSCT were followed for a median 5.3 years (interquartile range [IQR] = 4.6). For 65 participants with successful HSCT, BPF declined nonsignificantly (-0.055%/year; 95% CI -0.12, 0.006; p = 0.08). Twelve participants with graft failure had a 7-fold greater BPF decline (-0.42%/year; 95% CI -0.57, -0.26; p < 0.001; group difference p < 0.001). A smaller change in Hb after HSCT was associated with steeper BPF decline (0.035%/year per unit Hb; 95% CI 0.017, 0.053; interaction p < 0.001). WMH growth was similar in both groups (0.038% vs 0.030% per year; interaction p = 0.50) and was associated with older age and baseline WMHs, but not treatment success. DISCUSSION: Successful HSCT was associated with preserved brain volume, but not WMH growth, another brain health biomarker. Whether these HSCT-related brain structural changes translate to cognitive preservation remains to be determined. TRIAL REGISTRATION INFORMATION: ClinicalTrials.gov NCT02140554 registered 2014-05-14, first enrollment February 2015; ClinicalTrials.gov NCT00061568; registered 2004-07-16, first enrollment July 2004; ClinicalTrials.gov NCT02105766; registered 2014-04-01, first enrollment April 2014; ClinicalTrials.gov NCT03077542; registered 2017-03-10, first enrollment April 2017; ClinicalTrials.gov NCT00977691; registered 2009-12-14, first enrollment December 2009. CLASSIFICATION OF EVIDENCE: This study provides Class III evidence that in adults with sickle cell disease, successful hematopoietic stem cell transplant is associated with slower declines in BPFs over a median period of 5 years, compared with patients with transplant failure.
9 June 2026
Read appraisal →Physics in medicine and biology
Scan-wise generalized PET denoising with contrastive adversarial learning
Objective.Deep learning has significantly advanced low-count positron emission tomography (PET) denoising. However, models trained on specific distributions often yield biased outputs when applied to scans with different activity distributions caused by anatomical and physiological variations (distribution shifts). Existing methods fail to generalize well across these scan-wise variations. Our goal is to formulate PET denoising as a scan-wise domain generalization (DG) problem to mitigate these variations and achieve robust, unbiased denoising for unseen scans.Approach.We propose a contrastive adversarial DG (CADG) framework to learn scan-invariant features efficiently. We leverage the property that multiple low-count noise realizations can be generated from a single raw list-mode PET scan to form a scan-wise domain distribution. Unlike conventional adversarial training with cross-entropy (CE) loss, we propose a contrastive adversarial framework that minimizes the mutual information between feature and scan-wise domains. Considering that each subject can have multiple scans in longitudinal studies, we further propose an ordered and memory-queued contrastive adversarial framework. This method efficiently includes realizations from the same scan as positive pairs, different scans of the same subject as pseudo-positive pairs, and different subjects as negative pairs in a memory batch. We explicitly exploit their ordered relationship as prior knowledge using a novel noisy-robust multi-positive ordinal contrastive loss.Main results.We systematically validated the effectiveness of our methods using 1920 noise realizations derived from 80 subjects with 192 longitudinal scans of MK-6240 tau PET data. The proposed CADG approach demonstrated superior denoising performance compared to CE-based adversarial methods and standard baselines. The ordered contrastive loss successfully improved the peak signal-to-noise ratio and structural similarity index consistently while reducing bias and standard deviation in Alzheimer-related regions and the whole brain.Significance.To our knowledge, this is the first attempt to alleviate the performance degradation in cross-scan denoising from the perspective of DG. Our study is also a pioneering work in utilizing longitudinal scans as pseudo-positives within an ordered contrastive learning scheme to exploit fine-grained relationships for robust clinical PET imaging applications.
9 June 2026
Read appraisal →Blood advances
FDG-PET medullary total tumor volume highlights high-risk patients with newly diagnosed multiple myeloma in CASSIOPEIA trial.
This study aimed to assess the prognostic value of medullary total metabolic tumor volume (mTMTV) derived from fluorodeoxyglucose-positron emission tomography/computed tomography ([18F]FDG-PET/CT) compared with conventional PET-derived features and biological/chromosomal abnormalities in patients with newly diagnosed multiple myeloma (NDMM) treated with daratumumab for induction/consolidation and/or maintenance and enrolled in CASSIOPET, a companion study of CASSIOPEIA, with long-term follow-up. Automated bone/liver CT-based segmentation were applied to the baseline [18F]FDG-PET images, with mTMTV being defined using the median liver background as the cut-off, including focal lesions and diffuse bone marrow (BM) involvement. Both univariate/multivariate Cox and machine learning (ML)-based survival models were performed. A total of 195 patients were included, 81% of them PET-positive. Multivariate analysis demonstrated independent prognostic value of mTMTV for progression-free survival (PFS) (P< .001) and overall survival (OS) (P< .001), complementary to Revised International Staging System (R-ISS) (P = .008 and P< .001, respectively). The ML model confirmed these findings, achieving concordance index of 0.609 and 0.659 and identifying mTMTV as the most informative feature for PFS and OS. Adding R-ISS, BM maximum standardized uptake value (SUVmax) and anemia to mTMTV accounted for >60% of the ML model explanation for PFS and adding R-ISS, the number of focal lesions and BM SUVmax for >60% of the model for OS. Combining R-ISS and mTMTV enabled the creation of 2 new-risk subgroups. In conclusion, this prospective study demonstrated the prognostic relevance of [18F]FDG-PET/CT-based parameters in the initial workup of patients with NDMM in the era of anti-CD38-based therapy. mTMTV was found to have strong independent prognostic value, complementary to R-ISS and refining risk stratification. This trial was registered at www.clinicaltrials.gov as #NCT02541383.
9 June 2026
Read appraisal →Journal of robotic surgery
Technical challenges in autonomous robotic ultrasound examinations: perception, planning, and control
Ultrasound is one of the most widely used medical imaging modalities in clinical practice. Robot-assisted ultrasound systems (RAUS) offer significant advantages in reducing physician workload and improving imaging standardization. However, the dynamic and complex nature of the human physiological environment severely hinders the transition of RAUS from automation to full autonomy. This paper presents a narrative review of the core research progress of RAUS between 2021 and early 2026. It comprehensively analyzes the current technological landscape of autonomous robotic ultrasound by focusing on three core pillars: perception, planning, and control. By evaluating cutting-edge breakthroughs and existing limitations across these three domains, this paper identifies the key technical bottlenecks that restrict fully autonomous RAUS. Finally, we propose an evolutionary roadmap driven by multimodal fusion, digital twins, planning-control co-design, and embodied artificial intelligence (AI) architectures. This roadmap provides a developmental reference for autonomous robotic ultrasound examinations and also lays a crucial foundation for future autonomous ultrasound-guided surgical interventions.
1 June 2026
Read appraisal →Neurology
Associations of Baseline Clinical Phenotypes With White Matter Hyperintensity Volume Change: A Study of 4,329 UK Biobank Participants
BACKGROUND AND OBJECTIVES: White matter hyperintensities (WMHs) relate to cognitive and physical impairment. Although WMHs typically progress over time, regression has also been observed. Our aim was to explore baseline clinical phenotypes associated with subsequent WMH volume change in the UK Biobank (UKB). METHODS: We included participants with total volume of WMHs at first and follow-up brain MRI. We assessed 107 preselected clinical phenotypes measured at baseline and preprocessed using the Phenome Scan Analysis Tool (PHESANT). We derived a measure of WMH change as the residual from a linear regression of follow-up WMH volumes on baseline WMH volumes to account for baseline WMH burden. We ran (1) linear regressions for continuous WMH change as the outcome and (2) multinomial logistic regressions comparing progression, regression, and stable groups after categorizing WMH volume change using an established percentile-based approach. Models were unadjusted, partially adjusted (age, sex, total brain tissue volume, follow-up time), or fully adjusted (further adding blood pressure; BP). We corrected p values using false discovery rate (FDR) correction. RESULTS: We included 4,329 participants (median age 52; interquartile range 12), 54.6% female, with a median follow-up of 2.3 years (range 1-7 years). Across follow-up, total brain volume decreased (mean decrease 16,058 mm3) and WMH volume increased (median increase 293 mm3; both p < 0.001). WMHs progressed in 53.9% of participants, regressed in 26.01%, and remained stable in 20%. In fully adjusted linear models, higher diastolic BP was the only baseline phenotype associated with greater WMH progression (0.10 SD higher WMH change per 1 SD higher inverse-rank normal transformed diastolic BP; 95% CI 0.05-0.15, FDR p < 0.05). In multinomial models, hip pain lasting more than 3 months (odds ratio [OR] 0.25, 95% CI 0.11-0.58) and faster walking pace (OR 0.79, 95% CI 0.66-0.95) were inversely associated with WMH progression, whereas hip pain was also inversely associated with WMH regression vs stability (OR 0.31, 95% CI 0.12-0.79; all FDR p < 0.05). DISCUSSION: We established vascular risk factors associated with subsequent WMH volume change in the studied cohort. Distinct clinical and demographic profiles characterized WMH progression, regression, and stable groups. Results suggest that vascular factors relate to WMH change but are sensitive to covariate control. Further studies should establish factors differentially and causally related to WMH progression, regression, and stability.
1 June 2026
Read appraisal →Lancet (London, England)
Comparison of [18F]flortaucipir and [18F]MK6240 for the detection of tau pathology in Alzheimer's disease (HEAD): a multicentre, prospective, cross-sectional, within-participant study
BACKGROUND: Tau PET imaging has emerged as a critical biomarker for Alzheimer's disease, informing diagnosis, staging, and therapeutic selection. We investigated whether PET tracer selection alters tau detection. METHODS: We conducted a prospective, multicentre, non-randomised, within-participant comparison of [18F]flortaucipir (Tauvid), currently used in clinical settings in the USA and Europe, and [18F]MK6240, an investigational tau PET tracer. Participants were recruited from eight north American sites and underwent tau PET, amyloid-β (Aβ) PET, and detailed cognitive assessments. Tau PET with both agents was acquired within a 45-day window. Coprimary outcomes were the discriminative accuracy for Alzheimer's disease-related cognitive impairment and the frequency of tau positivity in early medial temporal lobe (MTL) and late neocortical regions. The study is registered with ClinicalTrials.gov, NCT05361382. FINDINGS: Between March 2, 2022, and Aug 27, 2025, 775 individuals were enrolled, with 682 completing all procedures (373 [55%] female, 309 [45%] male; 38 [6%] aged 19-27 years, 214 [31%] aged 50-65 years, and 430 [63%] aged 65-89 years). 32 (5%) participants identified as Hispanic or Latino. 637 (93%) identified as White, 24 (4%) as Black or African American, 16 (2%) as Asian, and five (1%) as other. In addition, 49 (7%) individuals were identified as being from a rural area. [18F]MK6240 showed greater accuracy than [18F]flortaucipir in distinguishing Alzheimer's disease from non-Alzheimer's disease impairment (area under the curve 0·93, 95% CI 0·89-0·95 vs 0·86, 0·75-0·91; p<0·0001). Among the older adults, tau positivity status was concordant in 560 (87%) for MTL and 603 (94%) for neocortical regions. In cognitively unimpaired participants, [18F]MK6240 identified twice as many MTL-positive cases as [18F]flortaucipir (n=54 [15%] vs n=23 [6%]). Prevalence ratio in Aβ-positive was 2·43 (95% CI 1·50-3·94; p=0·0003), identifying 23 additional cases per 100. Among discordant cases, 75 (89%) were [18F]MK6240-positive only and had higher Aβ burden (p<0·0001), APOEε4 frequency (p<0·0001), and cognitive impairment (p=0·0043) than those negative on both tracers. Neocortical tau positivity was more frequent with [18F]MK6240 than with [18F]flortaucipir in cognitively impaired individuals (80 [28%] vs 46 [16%]). Prevalence ratio in Aβ-positive was 1·74 (95% CI 1·32-2·29; p<0·0001), identifying 15 additional mild cognitive impairment and 21 dementia cases per 100. INTERPRETATION: Tau PET tracer selection influences the frequency of detection of tau pathology across the ageing and Alzheimer's disease spectrum. Compared with [18F]flortaucipir, [18F]MK6240 identified more individuals with tau pathology in cognitively unimpaired and cognitively impaired individuals, with direct implications for patient stratification in clinical trials and more precise guidance for therapeutic decision-making. FUNDING: National Institute on Aging.
31 May 2026
Read appraisal →PloS one
Three-dimensional shape cues affect human and artificial recognition systems differently
Humans and neural networks use shape and texture information differently. While humans weigh shape heavily in their ultimate classification decision, neural networks are more biased towards texture cues. Many tests of shape vs. texture bias have focused on shape recognition from an object's external contour. However, shape information is also conveyed through internal contours, shading, and attached shadows, especially when an object is viewed from noncanonical perspectives. Using models from ShapeNet, we created datasets of 120,000 texture-substituted images of objects from many viewpoints with and without shading and attached shadows. We tested humans' and several neural networks' ability to classify these objects by both their shape and their texture. Humans were much better at classifying texture-substituted objects by their shape than any network, although these differences were greater when shape was defined only by the external contour than when 3D cues were included. Our findings suggest that networks' texture bias is reduced when 3D cues are included in images. We next tested whether the inclusion of 3D cues benefitted humans and neural networks more for images of objects viewed from canonical or noncanonical perspectives. Consistent with earlier research, we found that 3D cues primarily benefitted humans for noncanonical images. For neural networks, the greatest performance gains were for canonical images. These findings suggest fundamental differences in how humans and networks use shading and attached shadows for object recognition. We argue that humans use these cues to infer objects' 3D structures while neural networks use them as another surface-level cue like texture.
31 May 2026
Read appraisal →European radiology experimental
Towards an accessible, centralised, searchable database for AI courses in Europe: the Artificial Intelligence in Medical Imaging and Radiation Oncology Education (AIMIROE) project.
OBJECTIVE: Artificial intelligence (AI) is transforming medical imaging and radiation oncology, yet limited understanding and access to education hinder adoption. This study, led by the European Society of Medical Imaging Informatics (EuSoMII) in collaboration with the European Federation of Radiographer Societies (EFRS), aimed to create an accessible, centralised, searchable database including all AI courses in Europe. MATERIALS AND METHODS: An electronic survey was developed to collect data on European AI course characteristics, such as format, delivery, content, target audience and European Qualifications Framework (EQF) level. This was disseminated via purposive sampling through social media and mailing lists of the EuSoMII and the EFRS between September 2024 and January 2025. Quantitative data were analysed using descriptive statistics and visual representations using Python Seaborn and Geopandas. RESULTS: This study identified 29 AI courses in Europe. Of them, 53.6% were offered by universities. Courses targeted radiographers (59%), medical physicists (52%), and radiologists (41%), mainly at EQF level 7 (44.4%). Most courses were standalone (65.6%) and online (55.1%), while 41.3% were free of charge. English was the primary language of delivery (79%). CONCLUSIONS: Different AI courses across Europe offer some entry-level knowledge but are often short in duration. Expanding formats, building practical competencies, providing multilingual access, and European-wide reach are essential for meaningful, practical, and equitable AI integration. RELEVANCE STATEMENT: With the scaling-up of AI adoption in medical imaging and radiation oncology, there is a variety of AI education provisions currently available. Accessing these options via an open, centralised, regularly updated database enables people to make an informed decision about their training and practise safely and meaningfully. KEY POINTS: We identified 29 different AI European courses varying in language, content, and delivery. Many clinical practitioners and researchers are unaware of these resources. We need a centralised database for customising AI learning choices and guiding future course design.
31 May 2026
Read appraisal →Journal of neuro-oncology
Predicting progression-free survival in glioblastoma with neuroimaging and machine learning
PURPOSE: Glioblastoma (GBM) is the most prevalent and aggressive form of malignant glioma. Reliable estimation of progression-free survival (PFS) prior to medical intervention could strengthen clinical decision-making and improve patient care. Here, we utilize machine learning (ML) to predict PFS in GBM patients using resting state network (RSN) connectivity before medical intervention. METHODS: GBM patients (N = 45, mean age 62.1 ± 10.3 years, mean PFS 9.5 ± 5.6 months, 62.2% male) were retrospectively recruited from Washington University Medical Center. All patients completed structural neuroimaging and resting-state functional MRI before surgery. Deep neural networks were trained on resting-state functional connectivity to predict PFS. Feature selection identified the 15 strongest predictive features prior to training. RESULTS: Sex (p = 0.0037), overall survival (p = 0.0003), MGMT promoter methylation status (p = 0.0064), presentation of weakness (p = 0.0037), and presentation of memory impairment (p = 0.045) were significantly associated with PFS. Tumor frequency and spatial correlation analyses associated dorsal attention, visual, frontal-parietal, and default mode networks with shorter PFS. Conversely, right-temporal lobe tumors were associated with better outcomes. RSN spatial maps revealed widespread alterations in association networks in GBM patients relative to controls. MRMR feature selection identified thalamic and association network connectivity, including somatomotor, ventral and dorsal attention, and default mode/parietal memory as the strongest predictors of PFS. Using leave-one-out validation, the model predicted PFS with an RMSE of 1.26 months, MAE of 1.08 months, and R² of 0.96 (p < 0.001). CONCLUSIONS: Our findings indicate that GBM alters functional brain organization on a widespread scale, and these global effects are informative of patient outcomes.
31 May 2026
Read appraisal →Journal of the Egyptian National Cancer Institute
Artificial intelligence in cancer immunotherapy: current trends in predicting response and personalizing treatment
Artificial intelligence (AI) can transform cancer immunotherapy by enabling more accurate prediction of treatment responses, the discovery of specific biomarkers, and the development of personalised treatment plans. Traditional single-marker biomarkers (PD-L1, TMB, MSI) lack consistency across tumour types and cannot be used to assess tumour heterogeneity or the dynamic tumour microenvironment (TME). This review synthesises developments in multimodal AI models that combine genomics, transcriptomics, radiomics, digital pathology (pathomics), circulating biomarkers, and clinical evidence to create composite predictive signatures with significantly better discriminatory value. AUCs over 0.8 have been seen in a few retrospective studies with deep learning and ensemble models on whole-slide images, CT/MRI/PET radiomics, spatial and single-cell omics, and multi-omics fusion models, but prospective and multicentre validation is scarce, and external validation often shows deterioration in performance. AI is also used to enhance the translational pipelines of adoptive cell therapies (e.g., CAR-T) by improving patient selection, manufacturing (e.g., digital twins), and early toxicity prediction (e.g., CRS, ICANS). Nevertheless, clinical implementation remains hindered by data heterogeneity, bias, poor longitudinal validation, limited reproducibility, and a lack of transparency in most models, even though prospective, multicenter validation and explainable AI are crucial for clinician trust and regulatory acceptance. New systems such as federated learning, foundation models, spatial omics, digital twins, and wearable monitoring represent paths to generalizable, privacy-preserving, and actionable systems in clinical practice. To achieve the potential of AI, the generation of data will need to be standardized, reporting must be transparent, interdisciplinary, and regulatory frameworks must be strengthened focusing on the practical use of AI and patient safety. By taking these steps, AI could be shifted to prospective clinical decision support, which uses AI to meaningfully enhance personalization and outcomes in cancer immunotherapy based on a retrospective research tool.
30 May 2026
Read appraisal →American Society of Clinical Oncology educational book. American Society of Clinical Oncology. Annual Meeting
Artificial Intelligence in Oncology: Clinical Applications, Challenges, and Opportunities
Artificial intelligence (AI) is reshaping cancer research and clinical oncology by enabling large-scale analysis of complex biomedical data. Although early AI efforts focused on single-modality tasks such as imaging interpretation or pathology classification, contemporary oncology increasingly requires integrated reasoning across heterogeneous inputs, including imaging, digital pathology, and genomics. Multimodal artificial intelligence (MMAI) models represent a major evolution by fusing these data types into unified predictive systems capable of informing diagnosis, risk stratification, treatment selection, and disease monitoring. This manuscript reviews the state of AI in oncology and highlights emerging foundation models within the field. There will be a focus on histopathology, and how this can be incorporated into risk assessment models. There will also be examples of MMAI utility in radiology and spatial proteomics. Additionally, we will discuss ethical and legal implications of MMAI in modern oncology, and recommendations for responsible application within clinical oncology practice.
30 May 2026
Read appraisal →European radiology experimental
Automated identification of MRI series using a hierarchical modular machine-learning pipeline
OBJECTIVE: The volume and diversity of large MR imaging datasets require efficient automated labelling tools for cataloguing MR series, as manual annotation is impractical and costly. However, relying on DICOM header fields alone is unreliable because sequence descriptors are heterogeneous and locally defined, frequently missing or incorrect, and may be altered or removed during anonymisation. MATERIALS AND METHODS: We developed an AI-based modular model to classify MR series. The pipeline comprises five sequential classifiers (Family, Weighting, Fat Suppression, Contrast, and Others) and was trained and tested on 18,181 MRI series from the multicentre PRIMAGE repository. The dataset was split by patient into 80% training/validation and 20% testing; within the training/validation subset, five-fold cross-validation was used. With the exception of contrast classification, all modules used DICOM tag-based machine learning models (CatBoost/Random Forest), while the Contrast classifier incorporated image analysis using a pretrained single-slice ResNet-50. Ethical approval for the study was obtained. RESULTS: Accuracy was 0.994 for Weighting, 0.984 for Family, 0.959 for Fat suppression and 0.958 for Others; the Contrast classifier recorded 0.841. Overall, the end-to-end classification yielded a weighted F1 of 0.849 (CI: 0.837-0.861) and an accuracy of 0.853 (CI: 0.841-0.865). CONCLUSION: The proposed approach provides a reliable and scalable solution for labelling large, heterogeneous MRI datasets across multiple anatomical regions. The pipeline achieved excellent performance for Weighting and Family classification, solid performance for Fat Suppression and 'Others'. However, Contrast classification remains the main limitation and warrants further refinement and/or additional modules. RELEVANCE STATEMENT: Reliable, automated MRI sequence labelling enables faster, reproducible cohort selection and protocol harmonisation in large archives, supporting downstream clinical research and AI tools while reducing manual curation effort and error. KEY POINTS: Multicentre PRIMAGE cohort (18,181 MR series) enabling evaluation in a heterogeneous setting. Machine learning model combining DICOM metadata and image features. High accuracy in series classification: up to 0.994 on key tasks. A scalable pipeline reduces manual annotation workload for radiologists.
30 May 2026
Read appraisal →Biomedical physics & engineering express
Biwt-UNet: lung nodule segmentation via wavelet transform and multi-scale feature fusion
Lung nodule segmentation is a crucial step in computer-aided diagnosis of lung diseases and is of great significance for early detection and diagnosis of lung cancer. Traditional segmentation methods face many challenges, such as blurred nodule boundaries that are difficult to distinguish effectively, and the diversity of nodule sizes and shapes, which together increases the difficulty of segmenting lung nodules. Existing approaches often fail to address these issues adequately, highlighting the need for more effective solutions. To meet these challenges, this paper proposes a hybrid lung nodule segmentation approach-Biwt-UNet that integrates improved Haar wavelet transform and multi-scale feature information from the multi-scale fusion module and bi-encoder fusion module proposed in this study. The approach enables accurate segmentation of lung nodules of varying sizes, diverse shapes and, in particular, indistinct boundaries. Experimental results on the public LIDC-IDRI dataset demonstrate that Biwt-UNet achieves excellent segmentation performance. It significantly surpasses other state-of-the-art methods with an average dice similarity coefficient of 90.18%, an average intersection over union of 82.67%, and an average normalized surface dice of 96.39%, fully validating the effectiveness and accuracy of the proposed model in CT-based lung nodule analysis. Moreover, ablation studies further confirm the individual contributions of each architectural component and show that the best performance is achieved within an acceptable parameter budget. This study provides a new technical perspective for automatic lung nodule segmentation and is expected to assist physicians in more efficient diagnosis.
29 May 2026
Read appraisal →Science advances
MR-AIV reveals in vivo brain-wide fluid flow with physics-informed AI
The circulation of cerebrospinal and interstitial fluid plays a vital role in clearing metabolic waste from the brain, and its disruption has been linked to neurological disorders. However, directly measuring brain-wide fluid transport, especially in the deep brain, has remained elusive. Here, we introduce magnetic resonance artificial intelligence velocimetry (MR-AIV), a framework featuring a specialized physics-informed architecture and optimization method that reconstructs three-dimensional fluid velocity fields from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). MR-AIV unveils brain-wide velocity maps while providing estimates of tissue permeability and pressure fields, quantities inaccessible to other methods. Applied to the brain, MR-AIV reveals a functional landscape of interstitial and perivascular flow, quantitatively distinguishing slow diffusion-driven transport [∼0.1 micrometers per second (μm/s)] from rapid advective flow (∼3 μm/s). This approach enables new investigations into brain clearance mechanisms and fluid dynamics in health and disease, with broad potential applications to other porous medium systems, from geophysics to tissue mechanics.
29 May 2026
Read appraisal →Scientific reports
Automated RECIST tumor response classification through prompt-guided large language models
This study investigates whether an entirely offline, general-purpose large language model (LLM) can reliably automate the classification of routine radiology reports according to RECIST (Response Evaluation Criteria in Solid Tumors) guidelines, focusing on how different prompting strategies support accurate, privacy-preserving tumor response assessment without additional model fine-tuning. An offline, in-house implementation of LLaMA-3.3 (70B) was used to classify real-world CT imaging reports from oncology patients. Reports were authored following RECIST-structured reporting but had outcome labels programmatically withheld prior to processing. Three prompting strategies-zero-shot, few-shot, and chain-of-thought prompting-were tested to guide the model in assigning RECIST categories: Baseline (BL), Complete Response (CR), Partial Response (PR), Stable Disease (SD), and Progressive Disease (PD). Model outputs were benchmarked against original expert labels using accuracy, precision, recall, and F1 scores. Across all tested prompting strategies, the LLaMA-3.3 model achieved strong classification performance. The best results were obtained with chain-of-thought prompting, reaching micro F1 scores of 0.81 across all RECIST categories. Overall model predictions aligned well with human expert assessments. Operating entirely offline within hospital infrastructure, the system preserved full compliance with stringent data privacy requirements. Prompt-driven large language models can accurately classify tumor response categories from real-world radiology reports in a scalable, reproducible, and privacy-preserving manner. Offline LLM deployment, combined with optimized prompting strategies, offers a promising approach for automating structured oncology report interpretation, potentially enhancing consistency and efficiency in clinical decision support workflows.
29 May 2026
Read appraisal →Journal of medical Internet research
Adaptive Fast-Slow Large Language Model Framework for Multidimensional Classification of Prenatal Ultrasound Reports: Comparative Study
BACKGROUND: Phenotype-driven prenatal diagnosis relies on the precise correlation between ultrasound findings and genetic outcomes; however, this process is hindered by the unstructured nature of clinical ultrasound reports. While large language models (LLMs) hold the potential to address this challenge, their specific application in this domain remains systematically underexplored. OBJECTIVE: To establish an effective LLM implementation framework for the clinical multidimensional classification of prenatal ultrasound reports, we evaluated the open-source DeepSeek-V3.2 family on real-world anomalous reports-covering both factual and subjective categories-while integrating retrieval-augmented generation (RAG) and chain-of-thought (CoT) reasoning. METHODS: From a cohort of 4256 pregnancies, we extracted 254 reports with fetal anomalies. We comprehensively evaluated both the high-speed base model (DeepSeek-V3.2-B) and the reasoning-enhanced model (DeepSeek-V3.2-R) across all 5 classification dimensions, comprising 4 factual extraction tasks-primary classification, standardized terminology, anatomical system, and abnormality count-and 1 subjective severity assessment. We further explicitly evaluated the efficacy of RAG for the subjective tasks. Finally, to validate the clinical utility of this approach, we performed a correlation analysis between the expert-validated multidimensional phenotypic profiles and definitive genetic outcomes derived from amniocentesis. RESULTS: While V3.2-B achieved high efficiency in factual tasks (accuracy and F1-score >90%), it underperformed in subjective severity grading (56.6% accuracy), exhibiting a recall of 0 for minor anomalies. Crucially, while RAG significantly improved both models' performance on internal retrieval datasets (P<.05), this benefit did not generalize to external test datasets (P>.05). In contrast, the V3.2-R model utilizing CoT reasoning achieved superior robustness (86% accuracy and F1-score=0.75) on external data without RAG; notably, introducing RAG to V3.2-R degraded performance to 81%, suggesting potential noise interference. Clinical validation against amniocentesis outcomes confirmed that accurate multidimensional phenotypic profiles significantly stratified pathogenic genetic risks. CONCLUSIONS: The rapid base models are efficient for factual classification, and RAG enhances performance on data similar to the knowledge base, whereas CoT is indispensable for subjective assessment. Within the constraints of our dataset and current retrieval implementation, CoT proved more robust than RAG for subjective assessment. However, this finding is specifically tied to our experimental setup and should not be generalized as a universal conclusion. We recommend clinically adopting this adaptive "fast-slow" LLM framework to efficiently perform the multidimensional classification of prenatal ultrasound anomalies. This privacy-preserving, locally deployable solution provides a scalable path to accelerate phenotype-genotype research and optimize invasive diagnostic decision-making.
29 May 2026
Read appraisal →European radiology experimental
Enhancing resolution and image quality in musculoskeletal MRI using deep learning reconstruction
OBJECTIVE: Deep learning-based noise reduction enhances image quality, overcoming the tradeoff among acquisition time, spatial resolution, and signal-to-noise ratio (SNR). We implemented deep learning reconstruction (DLR) into a 1.5-T musculoskeletal (MSK) magnetic resonance imaging (MRI) protocol to improve image quality without compromising SNR. MATERIALS AND METHODS: We retrospectively analyzed 39 MRI examinations performed on a 1.5-T scanner using standard-resolution (SR) sequences and sequences with higher resolution reconstructed with DLR (HR-DLR). Exams of the knees, shoulders, ankles, and hips were evaluated. The included sequences were: three-dimensional T2-weighted fast advanced spin-echo; T1-weighted and proton density-weighted fast spin-echo. One expert reader and two junior readers in agreement evaluated the visibility of various structures using a 5-point Likert scale in a blind manner. A fourth reader estimated the SNR and contrast-to-noise ratio (CNR) in bone and muscle. A mixed model was used to compare HR-DLR versus SR measures. The agreement between radiologists was assessed with the Kendall τ coefficient. RESULTS: The HR-DLR sequences globally had a smaller pixel size and shorter acquisition time. A good inter-reader agreement was obtained for SR sequences (0.613 ≤ τ ≤ 0.788) and even higher levels of agreement for HR-DLR sequences (0.682 ≤ τ ≤ 0.961). All the structures had higher or similar Likert scores in HR-DLR sequences (p < 0.001), regardless of joint and sequence contrast. Apparent SNR and CNR of HR-DLR and SR were similar. CONCLUSION: Incorporating DLR into 1.5-T MSK MRI protocols enhances resolution and maintains SNR and CNR, improving MSK structure visualization. RELEVANCE STATEMENT: This study demonstrated the effectiveness of deep learning reconstruction in improving the efficiency of 1.5-T musculoskeletal MRI exams. Despite shorter acquisition times, the visibility of key MSK structures was consistently rated as superior or similar in higher-resolution images with DLR. KEY POINTS: MRI is one of the primary diagnostic tools for evaluating MSK injuries and disorders. Deep learning reconstruction (DLR) implemented in a 1.5-T MSK protocol improved resolution, still preserving SNR and CNR. Higher scores were assigned by different raters to the DLR images, showing image quality improvement.
29 May 2026
Read appraisal →JMIR formative research
Virtual Reality-Based Training in Radiologic Technology for Contrast-Enhanced Computed Tomography Brain Imaging: Randomized Controlled Trial
BACKGROUND: Radiologic technology (RT) education faces challenges in bridging theory and practice due to limited clinical opportunities. While virtual reality (VR) enables safe and repeatable practice, a systematic instructional design framework is needed to develop scalable, procedure-focused modules. OBJECTIVE: This study evaluates the Radiologic Technology Virtual Reality (RTVR) framework that integrates 360-degree video capture, instructional overlays, interactive assets, and an immersive content authoring platform to deliver a contrast-enhanced computed tomography (CECT) brain scan module. METHODS: In this open-label, parallel-group, randomized controlled trial, 36 year-2 and year-3 RT students with no prior clinical training in diagnostic radiology at a university hospital in Thailand were randomly allocated (1:1) to a VR group or a conventional document-based instruction (control) group. The VR group completed the VR module, a grounded instructional design framework using 360-degree videos and a structured prebrief and debrief, for 20 minutes using a head-mounted display. The control group studied standard curriculum materials for the same duration. Blinding of participants was not possible. Outcome assessment was blinded. The primary outcome was declarative knowledge gain, assessed using a 20-item multiple-choice test before and after intervention. Secondary outcomes included technology acceptance, student satisfaction, and physiological responses during VR immersion. RESULTS: All 36 randomized participants (VR: n=18, control: n=18) completed the study and were included in the analysis. Experts validated the module as suitable and highly appropriate. Students reported high technology acceptance and satisfaction. Both VR and conventional methods produced substantial gains in declarative knowledge. No statistically significant difference in knowledge gain was detected between groups (test × group: unstandardized regression coefficient β=.056, 95% CI -1.360 to 1.473, P=.94). Year-2 students, who had less prior clinical exposure, showed larger pretest to posttest knowledge gains compared to year-3 students. Physiological monitoring showed a reduction in heart rate across the session, while blood pressure remained stable. No adverse events or VR-related discomfort requiring discontinuation was observed. CONCLUSIONS: The RTVR framework, which uses a real 360-degree video of authentic clinical settings, offers a scalable approach to procedural VR content creation without requiring specialist technical skills, distinguishing it from prior VR studies in radiography. These findings support the RTVR framework as a feasible, evidence-informed supplement to RT curricula for knowledge-focused procedural teaching, with learning outcomes comparable to those of conventional instruction in this context.
29 May 2026
Read appraisal →PloS one
Development and evaluation of AI model with deep learning for segmentation of extraocular muscles in thyroid eye disease
PURPOSE: To develop and evaluate an AI model for the segmentation of extraocular muscles (EOMs) using Magnetic Resonance Imaging (MRI). DESIGN: Single-center study with retrospective study. METHODS: The study included 52 patients with thyroid eye disease (TED) who underwent MRI examination of the orbital region at the Department of Ophthalmology, Osaka Metropolitan University, between October 2020 and June 2023. Manual labelling of all EOMs was performed on all slices. An AI model was created and compared with the manually labelled data using 5-fold cross validation on data sets of 12, 32, and 52 cases. The average signal intensity ratio (SIR) within the EOMs was measured. The primary outcome was the comparison of DICE similarity coefficients (DSC) as the agreement rate between AI's result and manual labelling amongst the three data sets (52 cases vs 32 cases vs 12 cases). The secondary outcome was the correlation of SIR between the AI and manual labelling. RESULTS: A significant difference in DSC was observed between the 12-, 32-, and 52-case data sets for the inferior rectus muscle and medial rectus muscle. A significant correlation was observed between the AI and manual labelling for SIR in all EOMs. CONCLUSIONS: An AI model was successfully developed for the automatic segmentation of EOMs. There was little measurement error between AI and manual labelling, and the AI model using 52 cases had improved measurement accuracy compared to the AI models using 32 and 12 cases.
29 May 2026
Read appraisal →JMIR medical informatics
Multimodal Prediction of Renal Tumor Malignancy From Radiology Reports and Structured Electronic Health Records: Retrospective Cohort Study
BACKGROUND: Accurate preoperative prediction of renal tumor malignancy is critical for guiding decisions and reducing overtreatment, as a substantial proportion of renal masses prove benign. Although radiology assessments and structured electronic health record (EHR) data are routinely used, many tumor-specific descriptors remain embedded in free-text radiology reports and are underused due to extraction challenges. OBJECTIVE: This study aimed to develop and evaluate a multimodal pipeline that integrates structured EHR variables with natural language processing features from computed tomography (CT) radiology reports, including large language model (LLM)-extracted abnormality characteristics and transformer-based report embeddings, to improve malignancy prediction. METHODS: We conducted a retrospective cohort study using University of Florida Health Integrated Data Repository Observational Medical Outcomes Partnership-mapped EHR data from December 2011 to August 2024. Adults with renal tumors were included if they had longitudinal diagnostic documentation consistent with a renal mass and at least 1 preoperative renal CT report; final benign or malignant status served as the outcome. Structured features included demographics, comorbidities, medications, vital signs, and laboratory measurements. From the recent preindex CT report, an on-premises LLM isolated kidney-specific findings and extracted abnormality characteristics. Four locally deployed LLMs were evaluated against manual annotations of 500 reports. Kidney-specific text was encoded using pretrained biomedical transformer models, including radiology Bidirectional Encoder Representations from Transformers (BERT) variants. We evaluated unimodal baselines and multimodal early, middle, and late fusion strategies. Model development used 5-fold cross-validation within the 80% training partition; each fold-specific model was evaluated on the same independent 20% held-out test set, with performance reported as mean and SD across the 5 held-out test evaluations. The primary metric was area under the receiver operating characteristic curve (AUC). RESULTS: The final cohort included 967 patients (n=712, 73.6% malignant). In extraction evaluation, Qwen2.5-32B achieved 88.3% overall accuracy with a 100% extraction success rate and was selected for downstream feature generation. Among unimodal models, the structured clinical variable model achieved an AUC of 0.758 (SD 0.012), kidney-specific text with radiology BERT achieved an AUC of 0.746 (SD 0.058), and abnormality characteristics alone achieved an AUC of 0.716 (SD 0.015). Multimodal fusion models achieved higher descriptive performance than unimodal models. Early fusion achieved the highest AUC (mean 0.813, SD 0.008), and F1-score (mean 0.809, SD 0.030), while late fusion achieved an AUC of 0.805 (SD 0.016). Ablation and interpretability analyses suggested complementary predictive information from structured clinical variables and kidney-specific text embeddings. CONCLUSIONS: Integrating unstructured radiology report text with structured EHR variables achieved higher mean predictive performance than unimodal approaches in descriptive comparisons. Multimodal fusion, particularly early fusion incorporating radiology BERT-derived kidney-specific text embeddings, achieved the strongest discrimination, suggesting potential value of natural language processing-enabled multimodal EHR pipelines for informing preoperative risk stratification.
28 May 2026
Read appraisal →European radiology experimental
Who labels best? Radiologists, rules, or large language models for CT reports on pulmonary embolism.
OBJECTIVE: To compare open-weight and proprietary large language models (LLMs), a rule-based extractor (RBE) and radiologists for labelling pulmonary embolism CT reports, and to test whether a hybrid RBE-LLM workflow improves labelling performance. MATERIALS AND METHODS: This single-centre retrospective study included structured CT reports from October 2021 to March 2025. Three labelling pipelines were evaluated: an RBE; a model-agnostic LLM extractor (18 open-weight, four GPT-4 variants); and a hybrid pipeline routing only RBE failures to an LLM. Ground truth was defined at the report-text level by deterministic schema matching for initially RBE-valid fields and blinded adjudication of RBE-invalid fields by two attending radiologists. Eight radiologists provided a human baseline. Outcomes included F1 scores, accuracy, LLM-based salvage of RBE failures, and labelling time. RESULTS: In total, 2,923 reports from 2,923 patients (mean age 66 ± 17 years; 1,465 women) were included. Falcon3-10b and GPT-4.1-mini achieved similar item-level performance (F1 0.98 [95% CI, 0.97-0.98] for both; p = 0.70) and both exceeded the RBE (F1 0.81 [95% CI, 0.80-0.82]; p < 0.001). Salvage of RBE failures was comparable between open-weight and proprietary models (88.1% vs 91.9%; p = 0.12). The hybrid RBE-LLM workflow achieved 99.8% accuracy and F1 0.99 (0.98-0.99), exceeding both the RBE and pooled radiologists (F1 0.92 [95% CI, 0.90-0.93]; all p < 0.001). CONCLUSION: Schema-constrained open-weight and proprietary LLMs exceeded rule-based extraction and, at the upper end of performance, matched a pooled radiologist label-transfer baseline. A rules-first, targeted LLM workflow enabled near-perfect extraction from finalised structured pulmonary embolism CT reports. RELEVANCE STATEMENT: A rules-first LLM workflow can automate high-fidelity extraction of structured CT findings from finalised radiology reports, enabling scalable, auditable, and more consistent cohort curation for clinical research, registries, and quality improvement. KEY POINTS: A hybrid rules-first workflow combining a rule-based extractor (RBE) with targeted large language model (LLM) salvage achieved the highest overall performance for labelling of pulmonary embolism CT reports (F1, 0.99; accuracy, 99.8%). The top standalone open-weight and proprietary LLMs (Falcon3-10b and GPT-4.1-mini) both exceeded the RBE and, at the upper end of performance, matched a pooled radiologist label-transfer baseline. The hybrid workflow reduced cohort-curation time from 32.2 h for radiologists to 1.0 h while reducing LLM calls by 85.6%, because the LLM was only triggered for rule-failed fields.
28 May 2026
Read appraisal →Journal of the American College of Cardiology
Opportunistic Cardiovascular Risk Assessment Using Routine Head CT in the Emergency Department
BACKGROUND: Routine noncardiac computed tomography (CT) imaging may contain information about cardiovascular risk. Head computed tomography (CTH) is among the most common imaging studies, conducted annually in millions of patients. Its utility for cardiovascular risk assessment has not been studied. OBJECTIVES: The purpose of this study was to develop and validate deep learning models for predicting incident cardiovascular disease (CVD) and estimating coronary artery calcium (CAC) scores from CTH, and assess performance against clinical risk factors. METHODS: This retrospective cohort study used data from the Stanford Health Care Emergency Department from August 2020 to August 2024. The CVD cohort comprised 27,990 adult patients without known CVD who underwent CTH. The CAC cohort included 2,313 patients who underwent both CTH and coronary CT angiography. Imaging features were extracted from CTH using pretrained deep learning models. Other risk factors were extracted from electronic health records. Outcomes were incident CVD complications (myocardial infarction, stroke, heart failure) and CAC scores (0, 1-10, 11-100, 101-400, >400 AU). Performance was evaluated using the concordance index (C-index) and area under the receiver operating characteristic curve, compared against the baseline model using the variables of the American Heart Association PREVENT (Predicting Risk of cardiovascular disease EVENTs) risk model. RESULTS: Four percent (1,110 of 27,990) of patients (median age 63.0 years, 51.7% female) experienced cardiovascular events. The CTH model achieved a C-index of 0.82 (95% CI: 0.78-0.85) compared with PREVENT (0.75; 95% CI: 0.70-0.79) with difference of 0.07 (95% CI: 0.04-0.10). For CAC estimation (n = 2,313, median age 65.0 years, 53.5% female), the CTH+PREVENT model achieved a C-index of 0.76 (95% CI: 0.72-0.80) and area under the receiver operating characteristic curve of 0.80 (95% CI: 0.73-0.85) for CAC >100. Of the patients, 15.7% were reclassified; higher-risk patients were younger but with higher prevalence of vascular calcifications (30.2% vs 24.8%, P = 0.001) and brain infarcts (20.1% vs 5.8%, P < 0.001). CONCLUSIONS: Routine CTH scans complement traditional risk factors for cardiovascular risk stratification, identifying subclinical disease in younger patients with favorable risk profiles. Clinical integration could improve CVD detection and prevention without additional costs or radiation.
28 May 2026
Read appraisal →Journal of radiation research
Feasibility of nonspinal bony landmark-based image registration in CyberKnife ® robotic radiotherapy for pelvic and femoral bone tumors
In stereotactic body radiotherapy for bone metastases using the CyberKnife system, spinal anatomy is generally employed as the alignment reference. However, when treating pelvic tumors such as those in the pubic bone or proximal femur, the spinal reference may be anatomically distant from the target, potentially compromising visualization and localization accuracy. This study evaluated the feasibility of using nonspinal bony landmarks, specifically the pubic bone and femoral neck, as reference points for image registration. Intrafractional positional deviations in six degrees of freedom and the frequency of intrafractional corrections were analyzed according to treatment site, including cases treated with conventional spine-based alignment. Alignment accuracy was visually and numerically verified by the treatment operator before beam delivery. Across all treatment sites, the mean translational and rotational deviations during treatment remained within ±0.5 mm and ± 0.5°, respectively. Furthermore, no substantial differences were observed in the mean frequency of intrafractional corrections among the different alignment approaches. These findings demonstrate that nonspinal tumor-adjacent bony landmarks can provide clinically acceptable accuracy comparable to conventional spine-based registration. To our knowledge, this is the first study to systematically assess the clinical feasibility of this approach in CyberKnife-based pelvic radiotherapy. The proposed method may serve as a practical and reliable alternative to spine-based alignment, particularly when the spine is remote from the target.
27 May 2026
Read appraisal →Journal of radiation research
Feasibility of retrieval-augmented generation for large language models with Japanese input in radiotherapy
Large language models (LLMs) have recently gained attention for their potential. However, concerns remain regarding their reliability due to limitations such as hallucinations and insufficient domain-specific knowledge. Retrieval-augmented generation (RAG) has emerged as a promising approach, enabling LLMs to reference external knowledge sources and generate accurate outputs. We aimed to clarify the potential of RAG-enhanced LLMs with Japanese input in the field of radiotherapy. This was assessed by evaluating performance on three certification examinations in Japan: the Japanese Medical Physicist Examination, the Japanese Board Examination for Radiologists, and the Japanese Board Examination for Radiation Oncologists. In this study, we constructed a RAG system named Rad-Hub, consisting of a Japanese Radiotherapy Knowledge Database (JRKD) and a retrieval framework built on Microsoft Azure. The JRKD was populated with 32 Japanese radiotherapy textbooks and clinical guidelines. We assessed its utility by inputting all multiple-choice questions from the three examinations into ChatGPT-4o, both with and without Rad-Hub, and recording the answers. They were then compared with reference answers determined by experienced medical physicists and radiation oncologists. Rad-Hub improved accuracy across all examinations. Accuracy increased from 77.0% ± 2.6% to 84.6% ± 1.5% in the Medical Physicist examination, from 74.9% ± 2.0% to 82.1% ± 1.1% in the Radiologist examination, and from 55.6% ± 4.4% to 71.7% ± 4.5% in the Radiation Oncologist examination. Performance gains ranged from 7.2% to 16.1%. These findings highlight the potential of RAG-enhanced LLMs, particularly ChatGPT-4o with Rad-Hub, for integration into radiotherapy applications, such as educational and clinical decision assistance.
27 May 2026
Read appraisal →Journal of breast imaging
Breast Arterial Calcification as a Predictor for Future Cardiovascular Events and Mortality: A Systematic Review and Meta-analysis
OBJECTIVE: Breast arterial calcification (BAC) is a surrogate marker of cardiovascular disease (CVD); however, its associations with incident mortality and major adverse cardiovascular events (MACEs) have not been systematically evaluated. The association between BAC and incident MACE and whether this is influenced by the method of BAC assessment was examined. METHODS: PubMed, MEDLINE, and EMBASE were reviewed to June 2025 for studies comparing BAC status with incident MACE (composite of death, myocardial infarction [MI], stroke, heart failure). The primary outcome was all-cause mortality. Random-effects meta-analysis was performed using hazard ratios (HRs) with 95% CIs. Subgroup analysis was conducted by method of BAC measurement and restricted to screening mammography cohorts, PROSPERO registration (CRD42020202612). RESULTS: Of 1118 studies screened, 24 were included (236 475 participants; mean age 58.5 years). Breast arterial calcification presence associated with all-cause mortality (HR, 1.32; 95% CI, 1.20-1.45; P <.0001) (follow-up 0.55-26 years). Increased risk was observed for cardiovascular mortality (HR, 1.38; 95% CI, 1.14-1.66; P = .0009), heart failure (HR, 1.29; 95% CI, 1.20-1.37; P <.0001), stroke (HR, 1.2; 95% CI, 1.11-1.30; P <.0001), and MI (HR, 1.2; 95% CI, 1.15-1.26; P <.0001). When restricted to screening mammography cohorts without concurrent indications for cardiovascular investigation (5 studies, n = 85 647), associations remained significant for all-cause mortality (HR, 1.34; 95% CI, 1.14-1.58; P = .0004), cardiovascular mortality (HR, 1.38; 95% CI, 1.14-1.66; P = .0009), heart failure (HR, 1.48; 95% CI, 1.19-1.86; P = .0005), and stroke (HR, 1.34; 95% CI, 1.08-1.67; P = .008). Associations with incident outcomes were observed only in studies using quantitative BAC assessment (HR, 1.31; 95% CI, 1.19-1.44; P = .002). CONCLUSION: Breast arterial calcification is a promising prognostic marker of mortality and CVD risk. Recognition of this incidental finding may enable timely identification of at-risk individuals.
27 May 2026
Read appraisal →Journal of perinatal medicine
Functional neurobehaviour of the human fetus: a comprehensive framework for prenatal assessment using 4D ultrasound and AI
OBJECTIVES: To evaluate current evidence on prenatal neurobehavioural assessment using four-dimensional ultrasound (4D-US), structured scoring systems, and emerging artificial intelligence (AI) platforms, and to develop an integrated framework for early identification of neurodevelopmental vulnerability. METHODS: A narrative review of the literature from 2000 to 2025 was conducted following PRISMA-guided organisational principles to enhance transparency. Studies were eligible if they used 3D/4D ultrasound to assess fetal neurobehaviour, including spontaneous motor activity, facial expressions, or behavioural transitions. Specific tools examined included the Kurjak Antenatal Neurodevelopmental Test (KANET), prenatal General Movements Assessment (GMA), and AI-assisted behavioural analysis systems. Methodological quality was appraised using the Newcastle-Ottawa Scale and the Joanna Briggs Institute checklist. Data extraction focused on imaging protocols, behavioural parameters, scoring systems, and associations with neonatal neurological outcomes. RESULTS: Fifty eligible studies demonstrated that fetal motor sequences, movement variability, and facial expressions exhibit hierarchical maturation consistent with developmental progression of brainstem, subcortical, and cortical neural circuits. KANET parameters showed reproducible scoring and meaningful correlation with neonatal neurodevelopment, particularly in high-risk pregnancies. Prenatal general movement patterns displayed continuity with postnatal repertoires and contributed to early neurological prediction. AI-based classifiers provided objective quantification of fetal movement and facial activity, supporting automated or semi-automated assessment workflows. CONCLUSIONS: Functional neurobehavioural assessment using 4D-US, structured scoring tools, and AI-enhanced analysis is feasible, reproducible, and clinically informative. Integrating behavioural markers with neurosonographic findings and computational modelling may strengthen early detection of neurological risk and improve long-term neurodevelopmental care pathways.
27 May 2026
Read appraisal →Journal of medical Internet research
Patients' Perspectives on the Implementation of AI in Radiological Diagnostics: Focus Group Study
BACKGROUND: Rapid developments in artificial intelligence (AI) will enable its widespread use in radiological diagnostics in the near future. Patients will then be confronted with findings generated with the help of AI. Understanding patients' perspectives on the use of this technology is one of the key factors for its successful implementation. OBJECTIVE: This qualitative study aimed to gain insight into patients' reasoning about the opportunities and risks of using AI in radiological diagnosis, to identify prerequisites for its acceptance, and to identify aspects that can promote trust in AI diagnoses, especially in scenarios of high personal concern. METHODS: A total of 7 focus groups were conducted with 34 patients (n=15, 44% female participants) aged between 23 and 85 years (mean 49.06, SD 17.08 y), recruited using purposive sampling strategies. Each focus group was audiotaped, transcribed, and analyzed using the method of structured qualitative content analysis. RESULTS: Study findings show that patients are open to the use of AI in radiological diagnostics. The basic prerequisites for this are (1) scientific evidence of safe outcomes that are more accurate and faster than those without AI; (2) recognizable added value in patient care; (3) transparency in the use of AI and disclosure to the patient; (4) comprehensive, binding measures for quality assurance; and (5) the use of AI solely to support the physician. However, the results indicate that further criteria are important for patients to be willing to choose a radiologist who uses AI and to trust AI diagnoses. In situations where they are personally affected, patients fear that physicians will place too much trust in the AI result and that the physician-patient relationship will become dehumanized. Therefore, the physicians' abilities and functions that inspire trust from the patients' perspective must come into play. These include (1) an independent diagnosis by the physician that takes into account not only the clinical context but also the individuality of the patient, (2) a comprehensible explanation of the pros and cons of using AI for patients and clear communication of AI output, and (3) a humane and empathetic physician-patient relationship, which shows that the physician continues to feel responsible for the patient. CONCLUSIONS: The results of the study underscore that a high quality of the entire "sociotechnical" system is an essential prerequisite for patient acceptance of the use of AI in radiological diagnostics and for trust in AI diagnoses. The further development of AI performance must go hand in hand with the creation of framework conditions for its use that meet patients' expectations of the role of the physician and ensure a trust-building physician-patient relationship. The study provides valuable insights into how such integration of AI into radiological practice can be achieved.
26 May 2026
Read appraisal →Rheumatology international
Diffuse splenic calcifications in systemic lupus erythematosus: a case-based review
Systemic lupus erythematosus (SLE) is a multi-organ autoimmune disease with a wide spectrum of splenic manifestations. In recent literature, splenic calcifications have increasingly been recognized as a rare manifestation of SLE. We describe a patient presenting with fever in whom abdominal imaging revealed splenomegaly with diffuse splenic microcalcifications, prompting further diagnostic evaluation that ultimately led to the diagnosis of SLE. To contextualize this finding, we performed a systematic case-based literature review and identified 29 previously reported cases, resulting in a total of 30 cases including our own. Across reported cases, no distinct serological or clinical phenotype could be identified. However, a characteristic imaging pattern emerged, predominantly consisting of diffuse, small, round-to-rod-shaped splenic calcifications. When longitudinal data were available, calcifications were usually stable, although progression was reported in association with disease flares, sometimes evolving to functional autosplenectomy. This case illustrates that splenic calcifications may represent an early or presenting manifestation of SLE and that an incidental detection should prompt consideration of autoimmune disease in the broad differential diagnosis. By synthesizing previously scattered reports, this review provides an updated overview to support clinical interpretation and awareness.
25 May 2026
Read appraisal →Neuroscience
MSCMH-Net: A multi-scale channel-mixing hybrid network for whole-brain segmentation
Whole-brain segmentation constitutes a fundamental task in medical image analysis, providing quantitative assessment of fine-grained brain regions and serving as a cornerstone for both clinical practice and neuroscience research. Despite its importance, the task is inherently challenging given the numerous brain regions, pronounced inter-class heterogeneity, and sophisticated inter-class spatial dependencies. Accurate whole-brain segmentation requires not only precise delineation of local features but also comprehensive modeling of long-range dependencies and global contextual information. To tackle these challenges, we propose the Multi-Scale Channel-Mixing Hybrid Network (MSCMH-Net), a CNN-MLP hybrid framework integrating convolutional and MLP modules at multiple hierarchical levels. The framework leverages the strengths of CNNs to capture local features and spatial structures, while employing MLPs to model long-range dependencies and global contextual information. For integrating global and local information, a channel-mixing module incorporating an exponential moving average (EMA) fusion strategy is employed. A composite dataset of 106 brain MR scans was used, including 36 from MICCAI-2012, 30 from ADNI and 40 from OASIS. Ground truth labels were annotated and double-checked by experts. Comprehensive experiments conducted on the composite dataset validate that MSCMH-Net achieves competitive results relative to existing approaches.
25 May 2026
Read appraisal →World journal of urology
From conventional imaging software to artificial intelligence: tools for stone volume assessment in urolithiasis. Α review by the EAU and YAU sections of endourology.
INTRODUCTION: Accurate assessment of stone burden is fundamental in urolithiasis, as it directly influences treatment selection and prognostic evaluation. Although maximum stone diameter on non-contrast computed tomography remains the most widely used parameter, it does not adequately reflect the three-dimensional complexity of urinary calculi. This review aimed to summarize the evolution of stone burden assessment from conventional imaging-based methods to software-assisted volumetry and artificial intelligence (AI)-driven tools, with emphasis on their accuracy, reproducibility, and clinical utility. METHODS: A narrative review of the literature was performed using PubMed/MEDLINE, Scopus, and Google Scholar for English-language studies published up to March 2026. Original studies, validation studies, technical reports, reviews, and guideline-related papers addressing conventional CT-based measurement, software-assisted volumetry, AI-based stone segmentation, and the clinical significance of stone volume were included. Due to heterogeneity in study design and reported outcomes, the evidence was synthesized narratively. RESULTS: Maximum stone diameter remains simple and widely available, but it incompletely represents true stone burden, particularly in larger or irregular stones. Formula-based ellipsoid calculations are practical yet show limited accuracy in complex geometries. Semi-automated CT-based segmentation software provides more reliable volumetric assessment, with excellent agreement with reference standards and reduced interobserver variability. AI-based approaches have further improved efficiency by enabling rapid and highly accurate automated stone detection and volume calculation. Across the reviewed literature, stone volume was consistently shown to be more clinically informative than linear dimensions for predicting spontaneous passage, stone-free rates after shockwave lithotripsy and ureteroscopy, and future symptomatic events during surveillance. CONCLUSIONS: Stone volume offers a more accurate and clinically meaningful estimate of stone burden than maximum stone diameter alone. The transition from formula-based methods to software-assisted and AI-driven volumetry represents an important advance in urolithiasis imaging. Wider adoption will depend on standardized imaging protocols, improved software accessibility, and validation of volume-based thresholds for routine clinical practice.
25 May 2026
Read appraisal →Science advances
Generalizability of an AI-based mammogram risk score (MRS) for breast cancer among diverse populations of women
Conventional prediction models incorporating genetic and clinical factors including breast density underperform in non-European populations. We investigated an artificial intelligence-derived mammogram risk score (MRS), a summary of texture features that captures intrinsic breast-tissue characteristics-the substrate for cancer development. This study leveraged data from two North American screening cohorts totaling >226,000 women, including non-Hispanic white, non-Hispanic Black, East Asian, South Asian, and Indigenous women. MRS distributions showed nonsignificant shifts (and similar SDs) between cohorts and across race and ethnic subgroups. MRS increased with age and was significantly associated with breast cancer risk, with hazard ratios per SD ranging from 2.24 [95% confidence interval (CI), 2.03 to 2.46] to 2.32 (95% CI, 2.25 to 2.39) after age adjustment. Associations remained significant within all subgroups. Calibration was excellent across the racial and ethnic groups and across full-field digital mammograms and tomosynthesis. These findings establish MRS as a strong predictor that is independent of race or ethnicity, demonstrating its potential for broader clinical utility.
24 May 2026
Read appraisal →Surgical and radiologic anatomy : SRA
Proximal femoral morphometry and its association with hip fracture patterns: a systematic review with meta-analysis
PURPOSES: Proximal femoral fractures (PFFs) represent a major cause of morbidity and mortality in the elderly population. Beyond bone mineral density, proximal femoral (PF) and hip joint (HJ) morphometry have been implicated in fracture susceptibility and fracture pattern. However, the relative contribution of individual geometric parameters remains incompletely defined. The purpose of this study was to systematically evaluate the association between PF and HJ morphometry and hip fracture (HF) patterns, and to assess the predictive value of radiographic morphometric parameters derived from standard anteroposterior radiographs. MATERIALS AND METHODS: A systematic review with meta-analysis was conducted in accordance with PRISMA guidelines. PubMed and MEDLINE databases were searched up to September 2025 for observational studies reporting radiographic morphometric parameters in patients with unilateral PFFs. Statistical meta-analysis was performed with random-effect models to compare intracapsular-ICF/subcapital-SCF fractures with intertrochanteric (ITF) and extracapsular (ECF) fracture patterns. RESULTS: Twenty-two studies comprising 4184 patients (77.8% female; pooled mean age 77.4 years) were included. Meta-analysis using a random-effects model revealed that ITF were significantly associated with a smaller femoral head diameter (p < 0.001) and increased medial neck cortex thickness (p = 0.021) compared with ICF. ECF demonstrated significantly longer femoral neck axis length (p = 0.012), increased horizontal offset (p = 0.048), smaller absolute offset (p = 0.026) and shorter hip axis length (p = 0.026) than ICFs. Angular parameters provided the most robust stratification. ITFs and ECFs exhibited significantly lower (more varus) neck-shaft angles (p < 0.001) and significantly higher Wiberg angles (p < 0.001) compared with the ICF group. CONCLUSIONS: PF and HJ morphometry are associated with distinct HF patterns, independently of bone mineral density. Simple linear and angular measurements obtained from standard anteroposterior radiographs may assist fracture pattern stratification, risk assessment, and preoperative planning in elderly patients.
22 May 2026
Read appraisal →Current obesity reports
Burden of Intrapancreatic Fat Deposition in Type 2 Diabetes and the Role of Obesity: A Systematic Review and Meta-Analysis
BACKGROUND: Growing evidence supports a role for high intrapancreatic fat deposition (IPFD) in the pathogenesis of type 2 diabetes mellitus (T2DM); however, the magnitude of this association and the extent to which it is influenced by body mass index (BMI), liver fat content (LFC), and age remain uncertain. OBJECTIVE: To quantitatively evaluate IPFD measured by magnetic resonance imaging (MRI) in individuals with T2DM and to investigate study-level factors contributing to between-study variability. METHODS: We systematically searched PubMed and Embase for observational studies comparing MRI-measured IPFD in individuals with T2DM versus non-diabetic controls, excluding studies in which IPFD quantification was performed using AI-based models. Pooled standardized mean differences (SMDs) were estimated using a restricted maximum likelihood approach with Hartung-Knapp adjustment. Heterogeneity was explored through subgroup, meta-regression, and sensitivity analyses. RESULTS: Thirty studies (n=3,980) were included. Individuals with T2DM had significantly higher IPFD than non-diabetic controls (SMD 1.13, 95% CI 0.74 to 1.51), with no significant differences when stratified by pancreatic region, MRI technique, geographic region, sample size, and risk of bias. In multivariable models, BMI remained the only consistent moderating covariate, and the excess IPFD in T2DM persisted after adjusting for BMI, LFC, and age (intercept=0.797, p=0.010, false discovery rate-adjusted p=0.018). Advanced heterogeneity exploration identified no structural outliers, and the pooled effect size remained consistent across multiple sensitivity analyses. CONCLUSION: T2DM is associated with substantially higher MRI-quantified IPFD, independent of study-level differences in BMI, LFC, and age. Although causality cannot be inferred, these findings support the hypothesis that IPFD may be involved in the underlying mechanism of T2DM. They also suggest that high IPFD could represent a metabolic hallmark, with potential implications for risk stratification and therapeutic targeting.
22 May 2026
Read appraisal →Open heart
Prognostic value of CT-derived fractional flow reserve for major adverse cardiovascular events in patients with suspected or known coronary artery disease: a systematic review and meta-analysis
BACKGROUND: Fractional flow reserve derived from CT (FFR-CT) enables non-invasive functional assessment of coronary stenoses in patients with suspected or known coronary artery disease, but evidence regarding its prognostic value remains fragmented. We conducted a systematic review and meta-analysis to quantify the association between abnormal FFR-CT and major adverse cardiovascular events (MACE), updating the 2022 meta-analysis by Nørgaard et al METHODS: We searched PubMed, Embase and Scopus through December 2025 for studies comparing outcomes in patients with suspected or known coronary artery disease with FFR-CT ≤0.80 versus >0.80 (PROSPERO: CRD420261276897). Random-effects meta-analysis with Hartung-Knapp-Sidik-Jonkman adjustment was performed. Subgroup analyses examined FFR-CT technology, geography and follow-up duration. Risk of bias was assessed using the Quality in Prognosis Studies (QUIPS) tool and certainty of evidence using Grading of Recommendations, Assessment, Development and Evaluations (GRADE) method. RESULTS: Twenty-two studies with 20 067 patients were included, representing a fourfold increase from the prior meta-analysis. Follow-up ranged from 12 to 120 months. Patients with FFR-CT ≤0.80 had significantly higher MACE risk (HR 3.94; 95% CI 2.92 to 5.31; p<0.0001) with substantial heterogeneity (I²=79.9%). However, restricting to hard endpoints (death/myocardial infarction; k=6) yielded HR 3.28 (95% CI 2.25 to 4.79) with zero heterogeneity (I²=0%). No significant differences emerged across FFR-CT technologies (p=0.812), including HeartFlow, machine learning and deep learning algorithms. Trim-and-fill analysis suggested possible publication bias, with adjusted HR 2.35. GRADE certainty was moderate. CONCLUSIONS: Abnormal FFR-CT is associated with nearly fourfold increased risk of MACE in patients with suspected or known coronary artery disease, with consistent prognostic value across technologies. The absence of heterogeneity for hard endpoints supports FFR-CT as a reliable prognostic marker. These findings address the evidence gap identified by current guidelines regarding FFR-CT prognostic utility. PROSPERO REGISTRATION NUMBER: CRD420261276897. CLINICAL TRIAL NUMBER: Not applicable (as this is a Systematic Review and Meta-Analysis and not a Clinical Trial).
22 May 2026
Read appraisal →Journal of medical Internet research
Explainable AI in Cancer Imaging: Scoping Review of Methods, Modalities, and Clinical Integration
BACKGROUND: Explainable artificial intelligence (xAI) is increasingly used in medical imaging to enhance transparency, clinical interpretability, and trust in artificial intelligence (AI)-assisted diagnostics, particularly in oncology. Evidence on how explainability is implemented, validated, and reported in cancer imaging remains fragmented. OBJECTIVE: This scoping review aimed to systematically map research applying xAI methods to radiologic cancer imaging, summarize methodological and clinical trends, and identify persistent gaps in validation and integration. METHODS: We conducted a structured search of PubMed and Scopus (final search executed on October 20, 2025), covering studies published from 2017 to December 2024. Eligible peer-reviewed articles using machine learning or deep learning were analyzed with a focus on xAI components. Data from 371 studies were extracted into predefined categories covering cancer type, imaging modality, AI model, xAI method, terminology, validation, code availability, and decision support system integration. RESULTS: Studies focused primarily on breast (112/371, 30.2%), lung (87/371, 23.5%), and brain (56/371, 15.1%) cancers, with prostate, thyroid, and liver cancers also represented. The primary imaging modalities were computed tomography (139/371, 37.5%) and magnetic resonance imaging (104/371, 28%). Deep learning was used in 70.1% (260/371) of studies, classical machine learning in 18.1% (67/371), hybrid pipeline methods for 10% (37/371), and emerging concept-, prototype-, or causal-based approaches accounted for 1.9% (7/371) of studies. Post hoc xAI methods were dominant (305/371, 82.2%), with visualization (163/371, 53.4%), and feature relevance (111/371, 36.4%) as the most common subcategories. Hybrid post hoc or inherent approaches comprised 12.1% (45/371) and intrinsically interpretable methods 5.7% (21/371). Data sources were mostly public (149/371, 40.2%) or mixed (100/371, 26.9%); 22.9% (85/371) used private institutional datasets, and 7.8% (29/371) did not report data sources. Among validated studies, expert or user-based validation was most common (104/193, 53.9%), followed by mixed methods (74/193, 38.3%), while quantitative metrics (10/193, 5.2%) and clinical knowledge-based (8/193, 4.1%) validation remained rare. Only 17.5% (65/371) of studies provided code and 12.1% (45/371) reported decision support system integration, with few achieving actual clinical deployment. CONCLUSIONS: This scoping review maps xAI implementation across multiple cancer imaging modalities, revealing methodological inconsistency and insufficient validation. Most research emphasizes visualization over quantitative interpretability, and few models are reproducible or clinically implemented. These findings provide an evidence base for researchers, clinicians, and regulators to prioritize standardization of xAI reporting, quantitative validation, and user-centered frameworks to advance trustworthy AI in oncology imaging.
22 May 2026
Read appraisal →European radiology experimental
MRI radiomics-based approach to predict pituitary neuroendocrine tumor invasiveness
OBJECTIVES: To assess the diagnostic potential of magnetic resonance imaging (MRI) radiomics and machine learning models using T2-weighted and contrast-enhanced (CE)-T1-weighted images, individually and combined, to predict the invasiveness of pituitary neuroendocrine tumors (PitNETs). MATERIALS AND METHODS: Patients with macro-PitNETs were retrospectively enrolled from 2019 to 2022. Radiomic features were extracted from manually segmented lesions on preoperative T2-weighted and CE-T1-weighted images and, after a feature selection step, used to assess invasiveness, defined following Trouillas' classification. Five machine learning models (logistic regression, random forest, gradient boosting, AdaBoost, XGBoost) were trained using CE-T1-weighted, T2-weighted, and CE-T1-weighted plus T2-weighted features. Performance was evaluated on a test set using the area under the receiver operating characteristic curve (AUC). RESULTS: Two hundred patients were included in the study: 95 PitNETs were noninvasive (74 grade 1a; 21 grade 1b) and 105 invasive (70 grade 2a; 35 grade 2b). A total of 102 radiomic features were extracted per sequence. The best-performing model was the XGBoost, using five combined CE-T1-weighted and T2-weighted features, with an AUC of 0.85 (95% confidence interval: 0.75‒0.95). Lower AUC values were obtained for logistic regression using CE-T1-weighted images (0.80) and AdaBoost using T2-weighted images (0.78). CONCLUSION: The XGBoost model, incorporating tumor shape, texture, and first-order features extracted from both CE-T1-weighted and T2-weighted MRI, showed high performance in predicting PitNETs invasiveness. This radiomic model might help identify tumors with a higher risk of disease persistence, recurrence, or progression. RELEVANCE STATEMENT: The radiomic model based on contrast-enhanced T1-weighted and T2-weighted MRI demonstrated high discriminative ability in predicting invasiveness of pituitary neuroendocrine tumors and could aid in identifying tumors that may be at higher risk for recurrence or progression, ultimately improving patient outcomes through personalized treatment strategies. KEY POINTS: Pituitary neuroendocrine tumors (PitNETs) represent a significant challenge in clinical practice. Accurate preoperative prediction of PitNET invasiveness is crucial for surgery and prognosis. Contrast-enhanced T1-weighted and T2-weighted MRI-based radiomic model effectively predicts PitNET invasiveness. The developed radiomic model could help optimize individualized treatment decisions before surgery.
21 May 2026
Read appraisal →European radiology experimental
Predicting resistance to neoadjuvant chemotherapy in osteosarcoma using machine learning with clinical data and T2-weighted MRI radiomics
OBJECTIVES: Identifying patients at risk of chemoresistant osteosarcoma enables risk-adapted management. This study aimed to predict chemoresistant osteosarcoma using baseline clinical and magnetic resonance (MRI)-derived radiomics features, with histological response as the reference standard and external validation included. MATERIALS AND METHODS: This retrospective single-center study included 115 patients with osteosarcoma from an institutional registry as the internal cohort, divided into training and test sets, and 49 patients from another institution as an external validation cohort. Tumor and peritumoral regions were manually segmented on baseline fat-suppressed T2-weighted MRI. Radiomics features were extracted using PyRadiomics, followed by two feature selection methods to identify potential predictors. Six machine learning models with varying feature combinations were trained to classify histologic chemoresistance in the internal training set. Model performance was assessed in the internal test set, and the best model was externally validated. RESULTS: The support vector machine model combining eight tumor radiomics features and four clinical-imaging parameters (presence of tumor necrosis > 50% on contrast-enhanced MRI, age, body mass index, and presence of metastasis at presentation) demonstrated the best performance. In the internal test set, it achieved a sensitivity of 83.3%, a specificity of 72.7%, an area under the receiver operating characteristic curve (AUROC) of 0.84, and a positive likelihood ratio of 3.06. External validation yielded a sensitivity of 88.5%, a specificity of 47.8%, and an AUROC of 0.77. CONCLUSION: A model combining tumor radiomics and clinical parameters at diagnosis showed strong performance in predicting chemoresistant osteosarcoma, with results confirmed by external validation. This approach may support personalized treatment strategies in high-grade osteosarcoma. RELEVANCE STATEMENT: The validated model may support early, individualized osteosarcoma management. KEY POINTS: Baseline T2-weighted MRI radiomics and clinical data can predict chemoresistant osteosarcoma. Tumor radiomics combined with clinical features achieved strong predictive accuracy. The support vector machine model reached an AUROC of 0.84 for the internal testing and of 0.77 for the external validation. The validated model may support early, individualized osteosarcoma management.
21 May 2026
Read appraisal →BMJ open
Evaluation of a deep learning model applied to chest X-rays of patients with suspected pneumonia presenting to the emergency department designed to predict admission risk: a retrospective training and prospective non-intervention validation.
OBJECTIVES: To develop, train and test a deep learning model Image-based PROgnostication applied to Chest X Rays (IPRO-X) tool that predicts the inpatient (IP) admission risk in patients with suspected pneumonia presenting to the emergency department (ED). DESIGN: The study consists of a retrospective (training) and prospective non-interventional shadow deployment (validation) of a deep learning model. SETTING: Three-hospital tertiary care system with two EDs. PARTICIPANTS: Consecutive adult patients (18 years old or older) who presented to the ED from December 2022 to February 2023 with a chief clinical complaint potentially related to pneumonia were reviewed for eligibility (n=5567), with a final number of 3677 included in the validation study. PRIMARY AND SECONDARY OUTCOME MEASURES: IPRO-X was developed using standard two-dimensional convolutional neural network InceptionNet architecture and processes chest radiographs (CXRs) acquired at admission generating a continuous value from 0 (IP-negative) to 1 (IP-positive). We examined IPRO-X's ability to predict IP admission using accuracy, area under the curve (AUC), receiver operating characteristic, sensitivity, specificity, positive predictive value and negative predictive value. IPRO-X scores were compared against observed outcomes (admission and discharge) to determine clinical utility in an ED setting. Four thresholds were defined from the retrospective phase: Youden optimal, highest specificity and thresholds equivalent to 30-day mortality rates of Pneumonia Severity Index (PSI) Risk class IV and Confusion, Respiratory Rate, Blood Pressure and Age 65 or older Score (CRB-65) scores 1 and 2. Performance was also analysed per group of chief complaints. RESULTS: In the validation set, 3677 patients (1777 (48%) female, median age 56 years (min 18, max 99)) were included. IPRO-X predicted IP admissions with an AUC of 0.795, and in the pneumonia-specific chief complaints, AUC increased to 0.828, whereas in non-related pneumonia chief complaints, the AUC decreased to 0.755. IPRO-X score was significantly higher in admitted patients compared with discharged patients (p<0.001). The accuracy of IPRO-X when using the PSI-anchored threshold was 0.751 and with the CRB-anchored threshold was 0.729, with a specificity of 0.803 and 0.968, respectively. CONCLUSIONS: IPRO-X applied to CXR of patients with signs and symptoms commonly related to pneumonia in the ED can accurately predict IP admission.
21 May 2026
Read appraisal →Current oncology reports
Impact of 18-FDG PET/CT on Multidisciplinary Decision-Making and Management in Abdominal Malignancies: A Systematic Review and Meta-Analysis
PURPOSE: To systematically review and assess the impact of 18 F-FDG PET/CT on multidisciplinary team (MDT) decision-making and management in adults with abdominal malignancies. MATERIALS & METHODS: We searched Ovid MEDLINE, Ovid Epub, Cochrane CENTRAL, EMBASE, and Scopus upto August 31, 2024, for studies including 18-FDG PET/CT and MDT management in abdominal malignancies in adult patients and describing pre- and post-PET/CT treatment plans or MDT decisions. Data extraction followed by independent two reviewer assessment using ROBINS-I framework. A random-effects meta-analysis of proportions was performed for studies reporting management change attributable to PET/CT. RESULTS: Fourteen studies met inclusion criteria, of which seven provided sufficient quantitative data for meta-analysis. A random-effects meta-analysis of the seven eligible quantitative studies showed that 21.8% of patients (95% CI: 10.6%-33.0%, p < 0.001; range: 8-36%) experienced a management change directly attributable to PET-based imaging. Between-study heterogeneity was substantial (I² ≈ 85%, p < 0.001). Reported management changes included cancellation or addition of planned surgery, modification of surgical extent, change in treatment intent between curative and palliative strategies, and alteration of systemic therapy or radiotherapy plans. Evidence on downstream outcomes such as survival, quality of life, and cost-effectiveness was limited and heterogeneous, precluding quantitative synthesis. CONCLUSION: PET/CT leads to a change in MDT management in roughly one in five patients with abdominal malignancies, often through re-staging and identification of occult metastases that alter treatment intent or surgical strategy. The magnitude of impact varies widely by clinical context and tumor type. 1. PET/CT identifies occult metastases and clarifies disease extent, allowing MDTs to avoid futile surgeries and appropriately shift between curative and palliative strategies. 2. Approximately one in five patients experience a meaningful change in management after PET/CT, demonstrating its significant influence on real-world treatment planning. 3. Incorporating PET/CT into MDT discussions enhances diagnostic confidence and supports more coordinated, evidence-based care across surgical, oncologic, and radiologic teams. By highlighting which patients benefit most, the findings support more efficient use of PET/CT, emphasizing the need for standardized outcome measures and prospective data to refine patient selection.
19 May 2026
Read appraisal →PloS one
A novel hybrid framework integrating GA-driven 3D ResUNetGAN for MRI brain tumor segmentation
Accurate brain tumor segmentation by multi-modal MRI is crucial for diagnosis, treatment planning, and prognostic assessment. This work proposes a novel hybrid architecture within a GA optimized by 3D residual U-Net (ResUNet) with generative adversarial network (GAN) for the accurate segmentation of tumor subregions, namely the tumor core (TC), whole tumor (WT), and enhancing tumor (ET), using a dataset collected from BraTS2023 (437 training and 188 validation). By using GAs, we enable the computer to automatically find the best model structure and training settings, which helps keep the model simple enough to avoid overfitting while being powerful enough to generalise well to new data. The GAN part of our method trains the model to draw more realistic and consistent boundaries around tumor regions in MRI scans, thereby improving the overall quality of the segmentation. In our experiments, this GA-ResUNetGAN approach performed better than several leading methods, such as 3D UNet, DynUNet, UNETR, SwinUNETR, and the standard ResUNet, especially in accurately identifying TC, WT, and ET parts-achieving Dice scores of 0.94, 0.96, and 0.91 on the validation set. In addition, K-Fold cross-validation was performed to ensure consistent performance across different data splits. While the method involves more processing work than some baseline techniques, it provides improved segmentation quality and boundary delineation on the evaluated dataset. These results demonstrate the advantages of combining evolutionary optimization with adversarial deep learning for challenging medical imaging tasks.
19 May 2026
Read appraisal →Physics in medicine and biology
Nonperiodic dynamic CT reconstruction using backward-warping implicit neural representation with diffeomorphism regularization
Objective.Motion artifacts remain a major obstacle in dynamic computed tomography (CT) reconstruction, particularly for nonperiodic rapid motion such as cardiac imaging in patients with fast or irregular heart rates, where ECG-gated approaches are unreliable. The extreme limited-angle problem arising from insufficient angular coverage within a single cardiac phase makes accurate reconstruction of fine cardiac structures especially challenging. This work aims to develop a self-supervised framework that recovers high-resolution dynamic CT images from nonperiodic motion scenarios without external training datasets.Approach.We propose BIRD, an implicit neural representation (INR) framework for nonperiodic dynamic CT reconstruction, with three components: (1) a dual-feature representation decomposing the dynamic scene into topology-preserving and free-form features, enabling modeling of both deformable motion and residual variations such as contrast kinetics; (2) a backward-warping deformation model enabling direct ray-based training, overcoming the resolution limitations of prior forward-warping INR approaches at clinically relevant sub-millimeter resolution; and (3) a diffeomorphism-based regularization enforcing anatomically plausible deformation vector fields through bidirectional inverse consistency, without restricting representational capacity.Main results.BIRD was validated on digital and physical cardiac phantoms and retrospective patient data. In the XCAT phantom study under severe limited-angle conditions (0.5 s rot-1, 120 bpm), BIRD improved peak signal-to-noise ratio over the next-best method by 1.43 dB on the right coronary artery and 1.21 dB on the right ventricle. Curved planar reformation confirmed improved vessel continuity and boundary sharpness. On real projection data from a physical cardiac phantom and a retrospective patient scan, BIRD reduced motion artifacts and improved depiction of cardiac and vascular structures compared with conventional reconstructions.Significance.The proposed framework performs self-supervised reconstruction of nonperiodic dynamic CT images solely from projection data. It offers potential clinical applications including non-ECG-gated cardiac imaging for patients with arrhythmia, cinematic image sequences, and retrospective motion artifact correction in conventional CT scans.
19 May 2026
Read appraisal →Computers in biology and medicine
AI unleashed: A meta-analysis transforming radiological insights in diagnosing abdominal infections
INTRODUCTION: Artificial intelligence (AI)-based imaging modalities are next-generation diagnostic devices for abdominal infections that promise to provide enhanced diagnostic speed and accuracy. This systematic review and meta-analysis critically analyze the diagnostic accuracy of AI-assisted imaging modalities, including for appendicitis, pneumoperitoneum, and cholecystitis, to present a balanced estimate of their clinical utility. METHODS: A systematic literature search was undertaken in PubMed, Scopus, and Cochrane databases to search for studies that assessed the diagnostic accuracy of AI-based imaging modalities for abdominal infections. Eleven studies were included based on the inclusion criteria, and data were pooled for analysis. Diagnostic performance was measured by estimating sensitivity, specificity, likelihood ratios, diagnostic odds ratio (DOR), and area under the curve (AUC) using a random-effects model. Subgroup analyses were done to investigate the effect of infection type on diagnostic accuracy. RESULTS: AI-assisted imaging demonstrated an overall sensitivity of 0.891 (95% CI: 0.824-0.944) and specificity of 0.860 (95% CI: 0.784-0.922) for diagnosing abdominal infections. Subgroup analysis revealed that AI-aided computed tomography (CT) exhibited a sensitivity of 0.902 (95% CI: 0.850-0.948), while ultrasound (US) showed a sensitivity of 0.864 (95% CI: 0.792-0.922). The highest AUCs were observed for pneumoperitoneum (0.985) and appendicitis (0.947), underscoring AI's robust diagnostic capabilities across multiple pathologies. CONCLUSION: AI-imaging modalities significantly enhance diagnostic accuracy for abdominal infections, particularly for appendicitis and pneumoperitoneum. This meta-analysis underscores AI's clinical potential, though future research should prioritize multicenter studies to validate AI models' generalizability and ensure consistent performance across diverse healthcare settings.
17 May 2026
Read appraisal →Computers in biology and medicine
A unified multimodal framework for chest X-ray retrieval and disease prediction for clinical decision support
Recent advances in medical imaging and natural language processing enable new opportunities for automated diagnostic support. Chest X-rays (CXRs) remain the most common imaging modality for screening pulmonary, cardiovascular, and systemic diseases; however, the growing volume of studies and free-text reports can overwhelm clinicians. We propose a unified multimodal retrieval and prediction framework that jointly leverages DICOM-format CXRs and radiology reports by projecting visual and textual features into a shared semantic space. Trained with contrastive and multi-label objectives, the system supports disease classification, case-based retrieval, and explainable AI. Experiments on the Open-I dataset demonstrate strong performance, achieving a Macro AUROC of 0.95 and Macro F1-score of 0.71 across 22 diagnostic categories. The retrieval module attains high ranking quality (MRR and nDCG >0.93) with sub-millisecond query latency. Quantitative explainability analysis further shows strong agreement between attention- and gradient-based attribution maps (Pearson ρ≈0.92), supporting trustworthy clinical decision-making.
17 May 2026
Read appraisal →Computers in biology and medicine
LADNET: An MRI-based deep learning approach for Alzheimer's disease detection.
Alzheimer's disease (AD) is considered a leading form of dementia, found in the majority of individuals, and it accounts for the highest prevalence of all the types of dementia. It is a gradually progressing condition which can begin with mild memory impairment and may lead to the inability to engage in meaningful conversations in the future. In addition to that, it is also capable of reacting appropriately to the environment around it. According to official data, there has been an increase in reported cases of AD-related mortality in recent years, along with rising rates of the disease itself. Therefore, it is beneficial for patients to be diagnosed with AD as early as possible in order to maximize their chance of survival. In medical field, deep learning (DL) is employed in many applications, especially the identification of Alzheimer's disease, has shown considerable success. Based on the fact that these methods are capable of acquiring and extracting features from extensive datasets autonomously, they are highly suitable for analyzing intricate medical images with the aid of these methods. Design the Lightweight Alzheimer's Disease Net (LADNET) model for the current task. Accurate detection of AD using magnetic resonance imaging (MRI)-based scans is presented in this work. The model classifies the three dementia types, namely Mild Dementia (MD), Moderate Dementia (MDD), and Very Mild Dementia (VMD), while also accurately categorizing the non-demented (ND) category based on the severity of the condition. We measure the performance of the model we recommend against a publicly available Kaggle dataset that contains over six thousand images from four different types of images. LADNET model achieves 99.4% accuracy and 99% AUC, outperforming existing methods. With a lightweight design of approximately 1.2 million parameters and fast inference (4.2 ms per image), LADNET demonstrates strong potential for practical clinical deployment, where it could help extract biomarker information from conventional MRIs, potentially reducing patient burden and diagnostic costs.
17 May 2026
Read appraisal →Biological psychiatry
Multimodal Representation Learning for Parsing Biological Heterogeneity in Psychiatric Neuroimaging
For decades, psychiatric neuroimaging has searched for biomarkers of depression and other disorders, but they remain elusive in clinical practice. While the last 5 years have seen rapid progress, other large-scale correlative studies have found only small, unreliable links between brain measures and clinical symptoms. Growing evidence suggests that such limitations are not just about sample size but depend critically on how models represent data. This review traces a recent shift away from univariate methods to multivariate/multiview approaches that learn more effective representations of biological and symptom measures by flexibly learning multimodal latent representations. First, we review how linear multiview embedding methods have revealed reproducible biological depression subtypes but do not perform well in small samples or samples enriched for mild symptoms. Then, we consider newer work exploring more sophisticated representations for neuroimaging data, including deep-learning and graph-based representations, and multimodal extensions that uncover complex latent patterns that single-modality studies miss. Then, we review recent developments in foundation models, which, once trained on large corpora, can "transfer learn" readily to small clinical cohorts, potentially bringing the advantages of large-scale learning to small, privacy-limited data. Finally, we highlight emerging representation tools that treat the brain as a dynamic, stateful multivariate process. Taken together, these advances point to a future in which the value of neuroimaging will be determined not only by ever-larger sample sizes but also by data quality and by how well our algorithms capture the distributed, multimodal, and evolving nature of psychiatric disorders.
17 May 2026
Read appraisal →European heart journal
Artificial intelligence-based quantification of breast arterial calcifications to predict cardiovascular morbidity and mortality
BACKGROUND AND AIMS: Women are underdiagnosed and undertreated for cardiovascular disease (CVD). Automatic quantification of breast arterial calcification (BAC) on screening mammography can identify women at risk for CVD. This study aimed to determine whether artificial intelligence-based automatic quantification of BAC from screening mammograms predicts CVD and mortality beyond PREVENT scores in a large, racially diverse, multi-institutional population. METHODS: This retrospective cohort study included 123 762 women from two healthcare systems who had screening mammograms. Breast arterial calcification was quantified using a transformer-based neural network for segmentation. Breast arterial calcification severity was categorized as zero (0 mm2), mild (>0-10 mm2), moderate (>10-25 mm2), and severe (>25 mm2). Kaplan-Meier analysis, Cox proportional hazards, and Fine-Gray competing event models were used to examine the association between BAC and major adverse cardiovascular events (MACE). RESULTS: Breast arterial calcification was detected in 16.1% (internal cohort) and 20.6% (external cohort) of women and provided significant prognostic value incremental to the PREVENT score. In PREVENT adjusted models, a clear dose-response was observed. Compared with zero BAC, mild [internal: hazard ratio (HR) 1.32, 95% confidence interval (CI) 1.10-1.59; external: HR 1.28, 95% CI 1.17-1.39], moderate (internal: HR 1.75, 95% CI 1.23-2.50; external: HR 1.79, 95% CI 1.55-2.06), and severe BAC (internal: HR 3.29, 95% CI 2.15-5.05; external: HR 2.80, 95% CI 2.36-3.32) were all prognostic for any MACE. Each 1 mm2 increase in BAC conferred an additional 2%-3% risk for MACE (P < .001). CONCLUSIONS: Automatically quantified BAC is an independent predictor of MACE and mortality, adding prognostic value to the PREVENT score. This approach may provide an opportunistic cardiovascular risk assessment during routine mammography screening without additional radiation exposure to guide earlier and more effective preventive care for women.
16 May 2026
Read appraisal →Physics in medicine and biology
ReaderAdaptNet: modeling reader variability in breast imaging with reader-specific embeddings
Objective.Inter-reader variability remains a major challenge in breast imaging interpretation, particularly for ordinal classification tasks such as breast density and background parenchymal enhancement (BPE). These visual assessments are prone to inconsistencies due to subjective interpretation, limiting the reliability of AI-based models trained on aggregated or noisy labels. This work aims to account for this variability rather than collapse it.Approach.We propose ReaderAdaptNet, a reader-adaptive network that explicitly models inter-reader variability through reader-specific embeddings. A novel two-stage deep learning framework underpins the training process of the model. The first stage jointly learns discriminative image features and low-dimensional embeddings representing individual annotation styles, enabling personalized classification. The second stage allows for embedding calibration using a small set of labeled examples, enabling rapid adaptation to new readers or consensus definitions without retraining the full model. The method is designed for multi-reader datasets and is evaluated on two breast classification tasks: breast density classification from full-field digital mammography and BPE classification from contrast-enhanced mammography.Main results.Results demonstrate that reader embeddings improve both individual and consensus-level performance, and calibrated embeddings enable flexible, low-cost personalization. Reader-specific embeddings of dimension 32 improved the mean classification accuracy across readers, increasing from 76.4% to 84.4% for breast density, and from 65.1% to 72.1% for BPE, compared to the proposed baseline method.Significance.By disentangling stable image features from reader-specific decision tendencies, ReaderAdaptNet provides a parameter-efficient, interpretable route topersonalization(alignment to a given reader) orunification(alignment to an institutional standard) under real-world variability.
16 May 2026
Read appraisal →BMC medical imaging
Development and validation of a novel deep learning coronary artery plaque quantification model
BACKGROUND: Coronary Computed Tomography Angiography (CCTA) is an established tool for assessing coronary artery disease. CCTA determined total coronary artery plaque volume predicts cardiovascular events, but manual quantification is impractical for routine use. Deep learning-based methods offer a promising solution for automated plaque volume assessment. METHODS: We developed and validated a novel software, QuantiPlaque, powered by deep learning models, for automated segmentation of total and calcified coronary plaque volume. Expert manual annotation served as the reference standard. Correlations between deep learning and expert segmentation were evaluated using intraclass correlation coefficient (ICC), Pearson's r, and Spearman's rho, and agreements were evaluated with Bland-Altman analyses. RESULTS: A total of 115 CCTA scans were included. Mean (range) age was 58 (51-64) and 51% were females. The mean coronary artery calcium score was 56 Agatston units, ranging from 0 to 601. The model demonstrated strong correlation and agreement with expert annotation for per-patient total plaque volume (ICC 0.95, mean difference - 8.35 mm3, 95% limit of agreement - 102 to 85 mm3, Spearman's rho 0.87, Pearson's r 0.95) and calcified plaque volume (ICC 0.90, mean difference - 0.28 mm3, 95% limits of agreement - 21.97 to 21.42 mm3, Spearman's rho 0.93, Pearson's r 0.90). Per-vessel analysis showed strong correlation in the left anterior descending artery and right coronary artery but was weaker in the left circumflex artery territory. CONCLUSION: The evaluated deep learning model provides accurate quantification of total and calcified plaque burden, with strong correlation and agreement to expert annotation.
16 May 2026
Read appraisal →Spine
Automated Classification of Cervical Spinal Stenosis Using Deep Learning on Computed Tomography Scans
STUDY DESIGN: Retrospective study. OBJECTIVE: To develop and validate a computed tomography-based deep learning (DL) model for diagnosing cervical spinal stenosis (CSS). BACKGROUND: Although magnetic resonance imaging (MRI) is widely used for diagnosing CSS, its inherent limitations, including prolonged scanning time, limited availability in resource-constrained settings, and contraindications for patients with metallic implants, make computed tomography (CT) a critical alternative in specific clinical scenarios. The development of CT-based DL models for CSS detection holds promise in transcending the diagnostic efficacy limitations of conventional CT imaging, thereby serving as an intelligent auxiliary tool to optimize health care resource allocation. MATERIALS AND METHODS: Paired CT/MRI images were collected. CT images were divided into training, validation, and test sets in an 8:1:1 ratio. The 2-stage model architecture employed: (1) A Faster R-CNN-based detection model for localization, annotation, and extraction of regions of interest (ROI), (2) Comparison of 16 Convolutional Neural Network (CNN) models for stenosis classification to select the best-performing model. The evaluation metrics included accuracy, F1-score, and Cohen κ coefficient, with comparisons made against diagnostic results from physicians with varying years of experience. RESULTS: In the multiclass classification task, 4 high-performing models (DL1-b0, DL2-121, DL3-101, and DL4-26d) achieved accuracies of 88.74%, 89.40%, 89.40%, and 88.08%, respectively. All models demonstrated >80% consistency with senior physicians and >70% consistency with junior physicians. In the binary classification task, the models achieved accuracies of 94.70%, 96.03%, 96.03%, and 94.70%, respectively. All 4 models demonstrated consistency rates slightly below 90% with junior physicians. However, when compared with senior physicians, 3 models (excluding DL4-26d) exhibited consistency rates exceeding 90%. CONCLUSIONS: The DL model developed in this study demonstrated high accuracy in CT image analysis of CSS, with a diagnostic performance comparable to that of senior physicians.
16 May 2026
Read appraisal →Computers in biology and medicine
Artificial intelligence for metabolic dysfunction-associated steatotic liver disease diagnosis: A systematic review
Metabolic dysfunction-associated steatotic liver disease (MASLD) poses significant diagnostic challenges. We conducted a systematic review of artificial intelligence (AI) applications in MASLD diagnosis, focusing on predictive modeling, screening, imaging analysis, and therapeutic assistance. Following PRISMA guidelines, we searched five databases (Web of Science, SciFinder, Elsevier ScienceDirect, EBSCOhost, Springer) for studies published January 2020-January 2026. Of 1247 initially identified records, 191 studies met inclusion criteria. AI models demonstrate high diagnostic performance across multiple modalities. Deep learning analysis of US, CT, and MRI enables automated steatosis quantification and fibrosis staging with accuracy comparable to expert interpretation. AI integration with multi-omics data facilitates novel biomarker discovery and therapeutic target identification. Despite these advances, model generalisability, standardization, and clinical integration remain challenges. AI holds substantial potential to transform MASLD diagnostics, but prospective validation and implementation strategies are needed before widespread clinical adoption.
16 May 2026
Read appraisal →The Journal of clinical ethics
Ethical Responsibility in the Off-Label Use of AI in Medical Imaging
AbstractArtificial intelligence in medical imaging (AI-MI), categorized by the U.S. Food and Drug Administration (FDA) as "Software as a Medical Device," can offer significant benefit to medical care. FDA approval-represented by a "label"-indicates that the device has demonstrated safety and efficacy for a specific clinical task and population for which the device is intended. Off-label use of AI-MI outside of the task and/or population can be medically and ethically problematic, yet benefits may be reduced or missed if entirely prohibited. Ethical responsibility frameworks can assist in evaluating and establishing boundaries and opportunities in off-label use of AI-MI. In this work, we review three different models for ethical responsibility and apply them to off-label use of AI-MI. We find that a future-looking collective responsibility framework, involving stakeholders such as industry developers, healthcare organizations, clinicians, and, most importantly, patients and/or their representatives, promotes proactive and practical actions to help ensure safe and ethical use of AI-MI off-label. By adopting this approach, stakeholders can balance innovation with safeguards, advancing patient-centered care while mitigating potential harms associated with off-label AI-MI use. This framework is crucial as AI-MI becomes increasingly integral to clinical decision-making processes.
15 May 2026
Read appraisal →European respiratory review : an official journal of the European Respiratory Society
Yield and safety of different biopsy types in interstitial lung disease: a systematic review and meta-analysis
BACKGROUND: Lung biopsy remains an important consideration in the management of interstitial lung disease (ILD). Transbronchial lung cryobiopsy (TBLC) offers an alternative to traditional surgical lung biopsy (SLB) and transbronchial forceps biopsy (TBB); however, up-to-date comparisons between these methods in ILD, particularly fibrotic ILD, remain limited. METHODS: A comprehensive literature search of Medline, Embase and Evidence-Based Medicine Reviews was performed for studies published between January 2000 and September 2025 to identify all citations reporting yield and complications of the three main lung biopsy methods for ILD. RESULTS: A total of 143 studies were included. Pooled multidisciplinary yield (following multidisciplinary discussion) in TBB, TBLC and SLB was 35.3% (95% CI 21.3-52.5), 84.3% (95% CI 81.7-86.5) and 88.6% (95% CI 83.4-92.3), respectively. For fibrotic ILD, the multidisciplinary yield was 12.5% (only one eligible study), 87.6% (95% CI 83.5-90.7) and 87.3% (95%CI 80.5-92.0), respectively. Meta-regression showed an increasing multidisciplinary yield of TBLC with more recent years of patient recruitment and an increasing percentage of fibrotic ILD in each study. The multidisciplinary yield of both TBLC and SLB was 85-90%, suggesting a biopsy "ceiling effect". All evaluated complications were more frequent in TBLC compared to TBB and some were more frequent in SLB compared to TBLC. Several complications were more frequent in fibrotic ILD compared to all ILD. CONCLUSIONS: The multidisciplinary yield of TBLC and SLB is equivalent, with both offering an acceptable safety profile. TBB has low yield, particularly in fibrotic ILD. These findings support using either TBLC or SLB as an initial diagnostic tool in fibrotic ILD, mainly based on patient-physician shared decision-making.
15 May 2026
Read appraisal →Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology
Metabolic, demographic, and behavioral risk factors predict brain structural variability in the general population: an analysis of the human connectome project young adults
BACKGROUND: Understanding whether commonly available metabolic, demographic, and behavioral factors can explain variability in brain structure may support the development of accessible predictive approaches. This study aims to evaluate the ability of machine learning regression models to predict brain structural measures in young, neurologically healthy adults using clinically accessible risk factors. METHODS: Data were drawn from the Human Connectome Project Young Adults dataset, including 1082 participants aged 22-37. Predictors comprised metabolic measures (systolic and diastolic blood pressure, hematocrit, body mass index, hemoglobin A1C, thyroid stimulating hormone) and demographic/behavioral variables (age, gender, race, education level, smoking history, alcohol use). Brain structural parameters from MRI included gray matter (GM) and white matter (WM) volumes, surface areas, and cortical thickness. Distance correlation (dcor) guided feature selection, followed by machine learning modeling. Model performance was evaluated using normalized mean absolute error (NMAE), normalized root-mean-squared error (NRMSE), and coefficient of determination (R²). SHAP values were used for feature interpretation. RESULTS: XGBoost consistently outperformed other algorithms, showing strong predictive accuracy for total brain segmentation, supratentorial, and total GM volumes (NMAE < 0.35, R² > 0.80). In contrast, cortical surface predictions (left superior frontal) showed moderate performance (NMAE > 0.50, R² < 0.60). SHAP analysis highlighted age, gender, education level, and diastolic and systolic blood pressure as the most influential predictors of brain structure. CONCLUSIONS: The findings highlight the methodological value of regression-based approaches for estimating brain structural variability and support their potential role as complementary tools for brain health research.
14 May 2026
Read appraisal →Hernia : the journal of hernias and abdominal wall surgery
Obturator hernia in adults: a systematic review of CT diagnosis and surgical management
PURPOSE: Obturator hernia is a rare but life-threatening pelvic floor hernia occurring predominantly in elderly, underweight women and is associated with delayed diagnosis and high perioperative mortality. This systematic review evaluated the diagnostic performance of computed tomography (CT), factors associated with bowel resection, and short-term outcomes of open versus laparoscopic repair in adults. METHODS: A systematic review was conducted per PRISMA 2020. PubMed/MEDLINE, Scopus, and Web of Science were searched from January 2004 to January 2026. Studies reporting at least three surgically confirmed adult cases were eligible. Three consecutive institutional cases were incorporated as an illustrative clinical adjunct. RESULTS: Twenty-two studies including 876 patients were analysed. The mean age across included studies was 79.7 years; 91.5% were female. Reported preoperative CT sensitivity across included studies was 93.2% (range 66.7-100%). Bowel resection was required in 31.6% and was associated with higher 30-day mortality (19.6% vs. 3.9%). Laparoscopic repair appeared to be associated with lower wound morbidity, shorter hospital stay (5.3 vs. 12.6 days), and lower recurrence (0% vs. 2.8%) in selected patient cohorts, although case-mix differences preclude direct comparison. Mesh was used in 44.3% overall. In three institutional cases (all elderly cachectic women presenting emergently), one underwent TAPP and two underwent open midline laparotomy; primary suture repair was performed in all owing to contamination, bowel compromise, or extreme frailty. CONCLUSION: CT remains the diagnostic cornerstone in obturator hernia. Surgical strategy should be individualised based on haemodynamic status, contamination, and frailty. Reported outcomes represent descriptive trends rather than pooled statistical estimates due to heterogeneity of available data.
14 May 2026
Read appraisal →Archives of gynecology and obstetrics
Clinical outcomes and practical management implications in rudimentary horn pregnancies: a systematic review and pooled analysis (2010-2025)
PURPOSE: Rudimentary horn pregnancies are rare and carry significant maternal and fetal risks. This study aims to systematically review published cases of rudimentary horn pregnancies describing data about time of diagnosis, time of rupture, and live birth rate (LBR), with exploratory comparisons between communicating (CHP) and non-communicating horn pregnancies (NCHP). METHODS: A PRISMA-guided systematic search of PubMed, MEDLINE, and Cochrane (October 1, 2025) was performed with terms "unicornuate uterus pregnancy" OR "rudimentary horn pregnancy." Case reports, series, and reviews with defined clinical data were included; abstracts and unclear reports were excluded. Three independent reviewers extracted data using a standardized protocol. Descriptive statistics (means, medians, standard deviations, and ranges) and exploratory group comparisons (Fisher's exact test, Mann-Whitney U and Student's t test with p < 0.05 denoting significance) were performed. RESULTS: From 132 articles, 190 cases were included: 27 CHP (14.2%) and 163 NCHP (85.8%). Time of diagnosis was earlier in NCHP (15.9 ± 8.9 weeks) than CHP (22.1 ± 10.0 weeks, p = 0.01). Rupture occurred in 68 cases (35.8%), similarly in CHP (48.1%) and NCHP (33.7%, p = 0.19). Time of rupture was earlier in NCHP (18.7 ± 6.6 weeks) than CHP (23.0 ± 9.2 weeks, p = 0.02). Four cases attempted to continue pregnancy; all resulted in premature rupture. Reported live births were infrequent (11.6%) and more common in published CHP (25.9%) vs. NCHP cases (9.2%; p = 0.02). CONCLUSIONS: This descriptive synthesis indicates uterine horn communication may be associated with later diagnosis/rupture and reported live births were significantly more frequent in published CHP vs. NCHP cases. However, all comparisons are exploratory and must be interpreted with extreme caution, as data limitations preclude causal inference. Exhaustive imaging and diagnosis are essential to characterize horn type. Surgical termination is advised, but if expectant management is chosen, close monitoring and multidisciplinary care, including weekly ultrasounds, are advisable to mitigate complications.
14 May 2026
Read appraisal →Scientific reports
Deep learning-based automatic field of view planning for prostate MRI in oblique coronal and oblique axial planes
Accurate field-of-view (FoV) prescription in oblique coronal and axial planes is essential for high-quality prostate MRI but remains operator-dependent and variable. We developed and evaluated a ResNet-based deep learning framework for automated FoV planning. In this retrospective multicenter study, FoV prescriptions were annotated on PI-CAI dataset. Three readers assessed intra- and inter-rater variability to establish reference consistency. Three neural network variants were trained on 1,474 examinations from PI-CAI dataset (2012-2021), and the optimal model was selected by internal validation. Generalizability and clinical utility were tested on three external cohorts totaling 530 examinations (2021-2024) using a non-inferiority design. The selected model achieved non-inferior performance for slice positioning, with differences ranging from 0.16 ± 0.99 to 0.37 ± 0.48. Across sites, FoV overlaps ranged from 82.4 ± 4.1% to 88.7 ± 6.0%, and the angle differences between predicted and reference planes were 4.66 ± 4.89° (Site I), 3.46 ± 2.80° (Site II), and 2.99 ± 2.90° (Site III). Clinical utility was high at all sites, with acceptability rates of 97.9%, 97.7%,98.8%, 98.1% and 98.1% for Site I (Raters 1-5), 95.7%, 97.8%, 100%, 95.7% and 97.8% for Site II (Raters 1-5), and 100% for all raters at Site III. These findings demonstrate the feasibility of automated FoV positioning for prostate MRI and indicate excellent clinical utility.
14 May 2026
Read appraisal →Scientific reports
Federated learning with swarm intelligence for efficient and secure medical image analysis
Collaborative learning in healthcare faces challenges, including strict regulations and fragmented data. This research introduces a federated learning framework that employs swarm intelligence to augment communication and enhance the analysis of medical images. The method optimizes hyperparameters, selects features, and assigns aggregation weights to federated clients simultaneously by combining Particle Swarm Optimization (PSO) and the Firefly Algorithm (FA) with deep Convolutional Neural Networks (CNNs). The framework was tested on three medical datasets: COVID-19 chest X-rays (5,856 images), monkeypox skin images (569 images), and breast cancer mammograms (320 images). These datasets were shared among four fake healthcare institutions. It strives to strike a balance between privacy, communication costs, and classification accuracy. The results showed that the test was 96.71% accurate in detecting COVID-19, 96.06% accurate in classifying monkeypox, and 97.0% accurate in diagnosing breast cancer. The framework was able to handle noise and attacks from individuals who sought to disrupt it, which reduced communication rounds by 25-30%. A privacy-utility analysis revealed that there were acceptable trade-offs, with accuracy remaining above 94%. This study employs robust privacy measures and statistical validation. It also shows how to use medical AI in smaller healthcare settings without putting patients' privacy at risk.
14 May 2026
Read appraisal →Pediatric surgery international
Selective non-operative management of post-appendectomy abscesses in children: a comprehensive systematic review and pairwise meta-analysis
Post-appendectomy intra-abdominal abscess remains a common complication following pediatric appendicitis. Although drainage has traditionally been considered standard treatment, increasing evidence suggests selected patients may be successfully managed with antibiotics alone. A systematic review and meta-analysis were conducted in accordance with PRISMA guidelines and registered in PROSPERO (CRD420251075191). PubMed, Embase, and the Cochrane Central Register of Controlled Trials were searched through June 20, 2025. Eligible studies included pediatric patients (0-18 years) with post-appendectomy abscess managed with antibiotic vs. invasive drainage. The primary outcome was failure of first-line treatment requiring escalation. Secondary outcomes included recurrence and length of hospital stay. Ten observational studies involving 363 pediatric patients were included (152 drainage, 211 conservative). Treatment success was 80.9% with drainage and 89.1% with antibiotics. Pooled analysis demonstrated no statistically significant difference in treatment success (RR 1.00; 95% CI 0.95-1.05) or recurrence (RR 0.99; 95% CI 0.57-1.74). Drainage was associated with longer hospital stays (MD 3.40 days; 95% CI 0.65-6.14; p = 0.02). Subgroup analyses demonstrated consistent findings. All included studies demonstrated a serious risk of bias due to confounding by indication. An antibiotics-first approach appears viable for clinically stable pediatric patients, achieving comparable outcomes with shorter hospitalization.
13 May 2026
Read appraisal →PloS one
Machine learning-based model for predicting recanalization in isolated distal deep vein thrombosis and analysis of predictors
BACKGROUND: Isolated distal deep vein thrombosis (IDDVT) is common, yet tools for predicting poor recanalization remain limited. We aimed to develop and compare machine learning models for predicting poor recanalization in patients with IDDVT and to identify the most informative predictors. METHODS: A total of 1600 patients with IDDVT were retrospectively enrolled. The dataset was randomly divided into a development set (n = 1280) and an independent test set (n = 320) using stratified sampling. Six predictive models were developed and compared: logistic regression (LR), support vector machine (SVM), random forest (RF), multilayer perceptron (MLP), extreme gradient boosting (XGBoost), and a Voting Ensemble. Model training and hyperparameter tuning were performed in the development set using five-fold stratified cross-validation, and optimal classification thresholds were determined using the Youden index. Model performance was evaluated by discrimination, calibration, and classification metrics, with 95% confidence intervals estimated by bootstrap resampling (10,000 iterations). SHAP analysis was applied to interpret the final model. RESULTS: In the independent test set, all models showed acceptable to strong discrimination, with AUC values ranging from 0.808 to 0.908. XGBoost achieved the best overall performance, with an optimal threshold of 0.183, an AUC of 0.908 (95% CI, 0.855-0.952), a Brier score of 0.077 (95% CI, 0.058-0.096), an accuracy of 0.900 (95% CI, 0.866-0.931), a precision of 0.650 (95% CI, 0.529-0.767), a recall of 0.803 (95% CI, 0.686-0.906), an F1-score of 0.717 (95% CI, 0.615-0.806), and a specificity of 0.918 (95% CI, 0.884-0.950). The calibration intercept and slope of the XGBoost model were 0.149 (95% CI, -0.192 to 0.454) and 1.410 (95% CI, 1.098-1.809), respectively, indicating acceptable overall calibration. SHAP analysis identified D-dimer rate, provoking-factor-related variables, anticoagulant use, and age group as the most influential predictors. CONCLUSION: Among six candidate models, XGBoost showed the best overall performance for predicting poor recanalization in patients with IDDVT. This study establishes an interpretable machine learning-based prediction framework focused specifically on poor recanalization in IDDVT and highlights the contribution of dynamic laboratory information, particularly D-dimer rate. The model may support early risk stratification and individualized follow-up planning, but external validation is required before routine clinical implementation.
13 May 2026
Read appraisal →European radiology experimental
Reconstruction algorithms and arm positioning effects on abdominal CT image quality and radiation dose: a phantom study
OBJECTIVE: To evaluate the effects of arm positioning and reconstruction algorithms on radiation dose and image quality of abdominal CT. MATERIALS AND METHODS: Abdominal CT scans were performed using a customized body phantom at 100 kVp with automatic tube current modulation based on noise indices (NI) of 9, 11, and 13 across seven arm positions: arms up (AU), arms down (AD), arms down with one or two-layered air cushions (ADAC, ADAC2), arms on the belly (AB), and arms on the belly with one or two-layered air cushions (ABAC, ABAC2). Images were reconstructed using filtered back projection (FBP), iterative reconstruction, and deep learning-based reconstruction (DLIR). Radiation dose was recorded. Quantitative image quality metrics included image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), sharpness, and structural similarity. Qualitative assessment of sharpness, noise, artifacts, and overall image quality was performed using fixed-dose AU-FBP images as the reference standard. RESULTS: Compared with the AU position, radiation dose increased by up to 83% in non-AU configurations, with similar trends across all NI settings. In contrast, air cushion use caused only minor, configuration-dependent dose changes. Qualitatively, non-AU FBP images showed marked degradation, whereas DLIR enabled selected non-AU configurations to approach near-reference diagnostic quality. Air cushion-related benefits were modest overall but more apparent in AB than AD configurations. CONCLUSION: Arm positioning is the primary determinant of radiation dose and image quality in abdominal CT. When arm elevation is not feasible, DLIR effectively mitigates image quality degradation, while air cushions provide limited, configuration-dependent benefits. RELEVANCE STATEMENT: When arms cannot be raised during abdominal CT, avoiding arms-on-belly and using air cushions may help maintain image quality without substantially increasing radiation dose, while deep learning image reconstruction further enhances diagnostic performance under restricted positioning conditions. KEY POINTS: Arm positioning is a major determinant of radiation dose and image quality in abdominal CT. Arms-on-belly positioning produces poorer image quality despite increased radiation dose. DLIR improves image quality across arm positions but does not eliminate positioning-related degradation.
12 May 2026
Read appraisal →Lasers in medical science
Machine learning enhanced optical spectroscopy for breast cancer diagnosis: A review
This literature review examines the transformative role of machine learning (ML) and deep learning (DL) in enhancing optical spectroscopy for breast cancer diagnosis. By synthesizing advancements from peer-reviewed studies (2015-2025), we evaluate how ML/DL integration improves the detection of malignancy-associated biochemical changes, enabling noninvasive, rapid, and accurate differentiation between healthy and cancerous tissues. This review highlights key spectroscopic modalities, such as Raman, fluorescence, diffusive optical spectroscopy (DOS), and photoacoustic spectroscopy (PAS), and their integration with AI-driven models, such as convolutional neural networks (CNNs), support vector machines (SVMs), and logistic regression. These techniques achieve diagnostic accuracies of up to 94% in subtype classification (e.g., luminal A, HER2-positive) by analyzing spectral biomarkers such as hemoglobin, lipids, and collagen. Challenges such as data variability, model interpretability, and clinical integration barriers are critically assessed. These findings underscore the potential of ML/DL-enhanced spectroscopy to standardize diagnostics, reduce unnecessary biopsies, and personalize treatment monitoring. Future directions emphasize the need for explainable AI (XAI), multimodal data fusion, and large-scale, diverse datasets to bridge translational gaps. By addressing technical, ethical, and regulatory hurdles, this integration promises to advance early detection, improve clinical outcomes, and reshape precision oncology.
11 May 2026
Read appraisal →AJNR. American journal of neuroradiology
Hippocampal Subfield Susceptibility Alterations in Mild Cognitive Impairment Revealed by 7T MRI
BACKGROUND AND PURPOSE: Hippocampal magnetic susceptibility alterations may serve as early neurodegenerative biomarkers of Alzheimer disease (AD), but subfield-level evidence in mild cognitive impairment (MCI) remains limited. This study uses submillimeter high-resolution quantitative susceptibility mapping (QSM) to investigate hippocampal subfield susceptibility changes in MCI. MATERIALS AND METHODS: Thirteen individuals with MCI and 13 cognitively normal controls (NC) underwent 7T MRI scans including MP2RAGE and multiecho gradient echo imaging (for QSM). The visibility of hippocampal strata radiatum, lacunosum, and moleculare (SRLM) was assessed using a visual rating scale. Hippocampal subfields (Cornu Ammonis 1 [CA1], CA2, CA3, dentate gyrus [DG], and subiculums) were segmented, and magnetic susceptibility values (χ) were extracted for each subfield. The susceptibility difference between the DG and other subfields was also calculated. Associations between subfield susceptibility measures and cognitive performance were then assessed. RESULTS: Qualitative analysis revealed a distinct hyperintense curved pattern in the hippocampus on the 7T QSM image, corresponding to the SRLM. In NC, this pattern was clearly visible exhibiting consistently delineated boundaries, while it appeared blurred or absent in most subjects with MCI. The visual rating results showed that NC had significantly higher scores than subjects with MCI (Hedgesg = 1.43, 95% confidence interval) [95% CI, 0.60-2.35; P = .003], with most NC rating 3-4 and most subjects with MCI rating 1-2. Quantitative analysis of hippocampal subfield susceptibility revealed significantly elevated susceptibility [χ] in the left DG in those with MCI compared with NC (Hedgeg = 0.88, 95% CI, 0.10-1.72; P = .03), leading to reduced susceptibility contrast between the SRLM and adjacent CA1-CA3 or DG. Furthermore, regional differences of susceptibility between key subfields (left CA1 versus DG and subiculum versus DG) were significantly reduced in MCI, reflecting a loss of subfield contrast. These quantitative changes were significantly associated with lower cognitive performance, even when controlling for age. CONCLUSIONS: 7T QSM reveals susceptibility changes in hippocampal substructures in individuals with MCI, providing a potential early biomarker for AD.
11 May 2026
Read appraisal →Biomedical physics & engineering express
Enhanced ResU-Net for brain tumor segmentation using EfficientNetB0, channel attention, and ASPP
Accurate and efficient segmentation of brain tumors is critical for diagnosis, treatment planning, and monitoring in clinical practice. In this study, we present an enhanced ResU-Net architecture for automatic brain tumor segmentation, integrating an EfficientNetB0 encoder, a channel attention mechanism, and an atrous spatial pyramid pooling (ASPP) module. The EfficientNetB0 encoder leverages pre-trained features to improve feature extraction efficiency, while the channel attention mechanism enhances the model's focus on tumor-relevant features. ASPP enables multi-scale contextual learning, which is crucial for handling tumors of varying sizes and shapes. The proposed model was evaluated on two benchmark datasets: The Cancer Genome Atlas Low Grade Glioma and brain tumor segmentation (BraTS-2020). Experimental results demonstrate that our method consistently outperforms the baseline ResU-Net and its EfficientNet variant, achieving dice similarity coefficient of 0.903 and 0.851, and HD95 scores of 9.43 and 3.54 for whole tumor and tumor core (TC) regions on the BraTS 2020 dataset, respectively. Compared to state-of-the-art methods, our approach shows competitive performance, particularly in whole tumor and TC segmentation. These results indicate that combining a powerful encoder with attention mechanisms and ASPP can significantly enhance BraTS performance. The proposed approach holds promise for further optimization and application in other medical image segmentation tasks.
11 May 2026
Read appraisal →AJNR. American journal of neuroradiology
High-Resolution 2D versus 3D Lumbar Spine MRI Optimized with a Deep Learning Reconstruction Algorithm and Prototype Conformal Coil
BACKGROUND AND PURPOSE: Lumbar spine MRI is predominantly performed using 2D FSE sequences. 3D FSE sequences offer potential advantages over 2D, especially with recent advances in deep-learning reconstruction (DLR) and the use of conformal high-density coil arrays to improve the SNR. This study aimed to compare image quality and diagnostically relevant differences between 2D and 3D lumbar spine protocols, both using a prototype conformal coil and DLR. MATERIALS AND METHODS: Subjects with lumbar curvature referred for routine lumbar spine MRI were prospectively recruited (n=31). A prototype, conformal spine coil array with 70 total posterior receiver elements was used at 3T. DLR was applied to both the 2D- and 3D-protocols to improve sharpness and reduce noise. For 3D, a prototype DLR ("Sonic DL") technique was applied to provide higher acceleration factors. Three attending musculoskeletal radiologists assessed 2D and 3D T1-weighted (T1), T2-weighted (T2), and STIR sequences for image quality and diagnostic performance. The Gwet agreement coefficient (AC1, AC2) or intraclass correlation (ICC) was used to assess interrater agreement, while the Wilcoxon signed rank test or a sign test with Bonferroni correction assessed differences between 2D and 3D sequences. RESULTS: There was an overall 14% scan time reduction of 3D compared with 2D sequences, and a 38% time reduction for T2 sequences. All 3D sequences exhibited similar image quality to 2D. 3D T2 sequences had less CSF flow artifact than 2D T2 sequences. Interrater agreement was mostly comparable between all 2D and 3D assessments, including excellent agreement for the spinal canal area (ICC = 0.952 [95% CI: 0.907-0.976] for 2D, 0.941 [0.866-0.973] for 3D), good agreement for foraminal stenosis assessments on T2 sequences (AC2 = 0.713 to 0.795 for 2D, 0.706-0.832 for 3D), and good agreement for fracture detection on T1 sequences (AC1 = 0.869 for 2D, 0.815 for 3D). CONCLUSIONS: DLR-enabled 3D lumbar spine MRI offers comparable image quality to 2D with reduced overall scan time.
11 May 2026
Read appraisal →PloS one
Reducing annotation burden in medical imaging with ADGNET: A semi-supervised deep learning strategy
We propose ADGNET, a semi-supervised framework for Alzheimer's disease (AD) diagnosis that jointly optimizes image reconstruction and classification through shared feature representations. The architecture integrates a residual backbone with attention modulation for dynamic feature selection, an encoder-decoder reconstruction branch for unsupervised representation learning, and a classification branch with focal loss to address class imbalance. This dual-task design enables effective feature learning from limited annotations. On two public MRI datasets-KACD (2D, 6,400 images) and ROAD (3D, 532 scans)-ADGNET achieves average performance improvements of 4.1% and 7.2% over state-of-the-art methods (ResNeXt WSL, SimCLR) across six metrics. Interpretability analysis using Grad-CAM and attention visualization confirms that the model focuses on clinically relevant neuroanatomical structures, particularly the hippocampus and temporal lobes, with strong correlation to established AD pathology (r = 0.67, p < 0.001). These results validate the model's exceptional generalization capability and feature representation effectiveness across multi-modal medical imaging data, offering an efficient solution for few-shot medical image analysis.
10 May 2026
Read appraisal →European journal of cancer (Oxford, England : 1990)
MuTriM: A multiscale deep learning model integrating longitudinal radiomics and pathomic features for predicting recurrence and adjuvant radiation benefit in breast cancer.
PURPOSE: We present MuTriM, a multimodal deep learning model integrating DCE-MRI and whole-slide pathology to predict survival and radiation benefit in breast cancer. PATIENTS AND METHODS: MuTriM employs an attention-based cross-modal and cross-temporal fusion framework that concurrently integrates DCE-MRI radiomic features with pathomic features of cellular morphology on WSI. MuTriM was trained on the Fudan University Shanghai Cancer Center (FUSCC) cohort (N = 335) and externally tested on the TCGA cohort (N = 126) using matched DCE-MRI and WSI. RESULTS: MuTriM was prognostic of recurrence-free survival (RFS) in a subtype-agnostic setting (HR = 5.26, 95% CI, 1.69-16.4, p = 0.004; C-index = 0.75), outperforming unimodal models based on DCE-MRI alone (C-index = 0.65) or WSI alone (C-index = 0.70). MuTriM was also prognostic among HER2 + (HR = 6.82, 95% CI, 1.08-43.2, p = 0.04) and ER+ populations (HR = 4.61, 95% CI, 1.25-17.1, p = 0.02). A high MuTriM score was associated with improved RFS (HR = 0.15, 95% CI 0.03-0.87; P = 0.03), whereas a low MuTriM score showed no significant benefit (HR = 4.06, 95% CI 0.05-31; P = 0.53) (interaction P = 0.04). Transcriptomic analyses revealed that high-risk tumors showed suppressed cytotoxic T-cell activity and enrichment of extracellular matrix remodeling, neutrophil degranulation, and neuronal signaling, consistent with immune evasion. Upregulation of NRXN1, GABRA3, KCNK3, and KLK12 was linked to worse survival, whereas SYNGR3 and CD207 were associated with favorable outcomes. CONCLUSION: By unifying radiology, pathology, and transcriptomics, MuTriM provides an interpretable multimodal framework for breast cancer prognosis and treatment guidance.
10 May 2026
Read appraisal →Diagnostic and interventional radiology (Ankara, Turkey)
Imaging in chronic thromboembolic pulmonary hypertension: review of the current literature
Chronic thromboembolic pulmonary hypertension (CTEPH) is a severe, life-threatening complication of pulmonary embolism with pulmonary hypertension (PH). The combination of insufficient resolution of thrombi following pulmonary emboli and accompanying microvascular disease results in PH. Advances in imaging can offer better insight into CTEPH diagnosis and management, but lack of disease awareness among radiologists has been shown to be a cause of CTEPH misdiagnosis or delayed diagnosis. This review highlights features pertinent to CTEPH diagnosis. The primary focus is on different modalities with their distinctive signs and newly developed technologies employing artificial intelligence systems.
10 May 2026
Read appraisal →Diagnostic and interventional radiology (Ankara, Turkey)
Foundation models for radiology: fundamentals, applications, opportunities, challenges, risks, and prospects.
Foundation models (FMs) represent a significant evolution in artificial intelligence (AI), impacting diverse fields. Within radiology, this evolution offers greater adaptability, multimodal integration, and improved generalizability compared with traditional narrow AI. Utilizing large-scale pre-training and efficient fine-tuning, FMs can support diverse applications, including image interpretation, report generation, integrative diagnostics combining imaging with clinical/laboratory data, and synthetic data creation, holding significant promise for advancements in precision medicine. However, clinical translation of FMs faces several substantial challenges. Key concerns include the inherent opacity of model decision-making processes, environmental and social sustainability issues, risks to data privacy, complex ethical considerations, such as bias and fairness, and navigating the uncertainty of regulatory frameworks. Moreover, rigorous validation is essential to address inherent stochasticity and the risk of hallucination. This international collaborative effort provides a comprehensive overview of the fundamentals, applications, opportunities, challenges, and prospects of FMs, aiming to guide their responsible and effective adoption in radiology and healthcare.
10 May 2026
Read appraisal →Current osteoporosis reports
The Use of Statistical Shape and Appearance Modelling for Predicting Osteoporotic Fracture Risk
PURPOSE OF REVIEW: Osteoporotic fractures remain a major cause of morbidity and mortality worldwide. Current clinical assessment metrics (e.g., bone mineral density) are limited in their ability to identify fracture risk and bone strength. Statistical Shape and Appearance Modelling (SSAM) offers a method to quantify anatomical geometry and density patterns. This review examines advancements in SSAM for osteoporosis research. RECENT FINDINGS: Recent literature demonstrates that SSAM can capture detailed bone geometry and internal density distribution. Increasingly, these models are combined with computational analytics, including finite element analysis and machine learning, to assess the mechanical and structural behavior of bone. SSAM provides a robust quantitative framework for bone research. Notably, SSAM is used to reconstruct 3D subjects from clinical 2D images for biomechanical evaluation. Although clinical adoption remains limited by generalizability, the advancement of deep learning and complex SSAM pipelines supports its potential for osteoporosis screening and fracture risk prediction.
10 May 2026
Read appraisal →Medical image analysis
Transfer learning from 2D natural images to 4D fMRI brain images via geometric mapping
Functional magnetic resonance imaging (fMRI) allows real-time observation of brain activity through blood oxygen level-dependent (BOLD) signals and is extensively used in studies related to sex classification, age estimation, behavioral measurements prediction, and mental disorder diagnosis. However, the application of deep learning techniques to brain fMRI analysis is hindered by the small sample size of fMRI datasets. Transfer learning offers a solution to this problem, but most existing approaches are designed for large-scale 2D natural images. The heterogeneity between 4D fMRI data and 2D natural images makes direct model transfer infeasible. This study proposes a novel geometric mapping-based fMRI transfer learning method that enables transfer learning from 2D natural images to 4D fMRI brain images, bridging the transfer learning gap between fMRI data and natural images. The proposed Multi-scale Multi-domain Feature Aggregation (MMFA) module extracts effective aggregated features and reduces the dimensionality of fMRI data to 3D space. By treating the cerebral cortex as a folded Riemannian manifold in 3D space and mapping it into 2D space using surface geometric mapping, we make the transfer learning from 2D natural images to 4D brain images possible. Moreover, the topological relationships of the cerebral cortex are maintained with our method, and calculations are performed along the Riemannian manifold of the brain, effectively addressing signal interference problems. The experimental results based on the Human Connectome Project (HCP) dataset demonstrate the effectiveness of the proposed method. Our method achieved state-of-the-art performance in sex classification, age estimation, and behavioral measurement prediction tasks. Moreover, we propose a cascaded transfer learning approach for depression diagnosis, and proved its effectiveness on 23 depression datasets. In summary, the proposed fMRI transfer learning method, which accounts for the structural characteristics of the brain, is promising for applying transfer learning from natural images to brain fMRI images, significantly enhancing the performance in various fMRI analysis tasks.
10 May 2026
Read appraisal →Medical physics
Spectral deep learning-based patient and bowtie scatter correction for clinical photon-counting CT
BACKGROUND: The presence of scatter in computed tomography degrades image quality, and can be caused by the patient and by other components in the beam path, such as the bowtie filter. While conventional energy-integrating detectors do not provide spectral distinction, photon-counting (PC) detectors are energy-selective and provide spectral information about the incoming X-ray photons. Since each energy threshold is affected differently by scatter, this spectral information implicitly encodes the scatter content of a projection. PURPOSE: The purpose of this work is to investigate how the spectral information can be exploited to improve deep learning (DL)-based scatter correction. Furthermore, the performance of joint and separate patient and bowtie scatter correction will be investigated, addressing that bowtie scatter has not been considered in current DL-based approaches. METHODS: We present a DL-based approach that can estimate bowtie and patient scatter jointly and compare it against a separate correction. We also introduce neural network-based methods that incorporate the spectral information inherent in PCCT for scatter correction. We present networks that estimate scatter for up to four energy thresholds simultaneously. Training and validation was performed with Monte Carlo data as well as with real data measured by a clinical PCCT system. RESULTS: When comparing joint and separate patient and bowtie scatter estimation, both methods reduce the mean absolute error (MAE) from 8 HU to 1 HU. All proposed DSE methods effectively reduce scatter artifacts and perform better than the convolution-based reference approach. Incorporating the spectral information further improves the performance, with the DSE variant with four energy thresholds achieving the best overall results for all thresholds. For all energy thresholds tested, the spectral DSE methods reduced scatter errors originating from the patient and the bowtie in PCCT from up to 8 HU to below 1 HU. In addition to the global MAE, we report a critical MAE (MAE10) restricted to voxels with uncorrected errors > $>$ 10 HU, as such deviations are visually perceptible in soft tissue and exceed the noise level of modern CT systems. In all test cases, the proposed spectral methods reduced the MAE10 from 23.8 HU in the uncorrected images to 1.6 HU after spectral correction. The affected voxels comprised on average 25 % of the image volume, indicating a significant reduction in artifact intensity in the most affected areas. In virtual monoenergetic images (VMI), the application of spectral neural networks resulted in a significant reduction in MAE from ∼ $\sim$ 16 HU to ∼ $\sim$ 2 HU at 45 keV, from ∼ $\sim$ 8 HU to ∼ $\sim$ 1 HU at 70 keV, and from 5 HU to under 1 HU at 100 keV. CONCLUSIONS: This paper presents a combined method for correcting patient and bowtie scatter that delivers results equivalent to separate corrections and thus eliminates the need for multiple networks. Further, we demonstrate that deep scatter estimation can effectively exploit the spectral information available to improve scatter correction, especially for spectral applications, like VMIs. Spectral DSE networks slightly outperformed non-spectral variants, with multiple energy thresholds lead to more accurate estimations. This enables the use of one network for scatter correction and eliminates the need for multiple ones, thereby saving computational cost and complexity.
10 May 2026
Read appraisal →Medical image analysis
Context-enriched contrastive auto-encoder with topology learning for medical hyperspectral image classification to diagnose tumors
Deep learning has emerged as a highly effective approach for the automatic classification of medical hyperspectral images (MedHSIs), facilitating the accurate diagnosis of diverse tumors. Most of current methods suffer from the challenges in the separability of latent representation and the preservation of topological structure. To remedy such deficiencies, we propose a novel context-enriched contrastive auto-encoder with topology learning (CCAET) based MedHSI classification for tumor diagnosis. Concretely, we initially employ an auto-encoder (AE) network as the foundational framework to achieve latent representation and data reconstruction. Subsequently, to enhance the separability of the latent representation, we integrate a context-enriched contrastive loss into the AE framework, where both individual pixels and their contextual information are used to redefine contrastive learning, so that small distances between intra-class features and large distances between inter-class features are achieved. Further, we construct a topology-preserved loss by incorporating a graph based approach into the foundational framework, which aims to maintain the topological structure during latent representation learning. Furthermore, we devise a gradient descent-based optimization scheme to optimize the overall loss of the proposed CCAET method, so as to obtain the final latent representation of MedHSI. Finally, we implement soft-max-based label prediction to obtain the classification results. Different from most existing methods, the CCAET approach considers both the separability of latent representations and the preservation of topological structure to enhance MedHSI classification performance. Extensive experimental evidence has been presented to substantiate that the proposed CCAET method outperforms several state-of-the-art techniques in terms of MedHSI classification based tumor diagnosis.
9 May 2026
Read appraisal →Artificial intelligence in medicine
Learning with less: A survey of deep learning in medical imaging under varying supervision levels
The rapid evolution of deep learning has significantly advanced the field of medical image analysis. However, despite these achievements, further progress is hindered by the scarcity of large, well-annotated datasets. To overcome this limitation, recent years have seen a growing focus on developing methods that leverage varying levels of supervision as alternatives to fully supervised learning. Instead of relying solely on complete annotations, these approaches employ partial annotations or other supervision strategies to train deep learning models. This paper presents a systematic review of deep learning methods for medical image analysis under different levels of supervision. To this end, we categorize these methods based on the level of supervision they rely on, encompassing categories such as no supervision, inexact supervision, incomplete supervision, inaccurate supervision, and only limited supervision. We further divide these categories into finer subcategories. For example, we categorize inexact supervision into multiple instance learning and learning with weak annotations. Similarly, we categorize incomplete supervision into semi-supervised learning, active learning, and domain-adaptive learning and so on. Furthermore, we systematically summarize commonly used datasets for different types of supervision in deep learning for medical image analysis and explore future research directions to conclude this survey.
9 May 2026
Read appraisal →Journal of clinical ultrasound : JCU
Integrative Machine Learning Model Leveraging DCE-MRI and PSA Values for Advanced Risk Stratification in Prostate Cancer
PURPOSE: Accurate grading of prostate cancer is critical for treatment strategies and risk stratification. This study aims to develop a machine learning (ML) model integrating Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) pharmacokinetic parameters with Prostate-Specific Antigen (PSA) values to predict ISUP grade metastatic risk groups. METHODS: This retrospective study included 102 patients with histologically confirmed prostate cancer. DCE-MRI pharmacokinetic parameters (Ktrans, Kep, Ve, CER, MaxSlope, IAUGC) were standardized. The dataset was balanced using the Synthetic Minority Oversampling Technique and split into training, validation, and test sets. ML models, including Random Forest, were evaluated using Area Under the Curve (AUC) values. RESULTS: The Random Forest classifier achieved the highest performance, with an AUC of 0.92. Precision-recall analysis identified an optimal threshold of 0.3, balancing sensitivity and specificity for high-risk group detection. SHAP analysis highlighted PSA, MaxSlope, and Kep as key predictors contributing to model accuracy. CONCLUSION: Integrating DCE-MRI parameters with PSA values using ML algorithms enhances the prediction of ISUP grade metastatic risk groups. This method provides a robust tool for metastasis screening and personalized treatment in prostate cancer.
9 May 2026
Read appraisal →Scientific reports
From intelligent models to clinical tools: the evolving landscape of AI in medical imaging
Integration of Artificial Intelligence (AI), particularly deep learning, into medical imaging represents a profound shift in diagnostic medicine, moving from purely descriptive analysis to advanced predictive and prescriptive analytics. This Collection explores the rapid advancement of AI-driven tools in their specific fields such as oncology, cardiology, ophthalmology and so on, highlighting their potential to improve diagnostic accuracy, workflow efficiency, and personalized treatment planning. However, significant challenges remain, including the heterogeneity of medical image data, the "black box" nature of some intelligent models, and the critical hurdles of clinical integration and validation. The research presented here addresses these frontiers, showcasing innovations in algorithm development, explainable AI, and translational application. This Editorial synthesizes the contributions and outlines the essential collaborative pathway-uniting computer scientists, clinicians, and regulatory bodies-required to translate algorithmic promise into robust, trustworthy, and equitable clinical tools that genuinely improve patient care.
9 May 2026
Read appraisal →Medical image analysis
Diversity-driven MG-MAE: Multi-granularity representation learning for non-salient object segmentation
Masked Autoencoders (MAEs) have grown increasingly prominent as a powerful self-supervised learning paradigm. They are capable of effectively leveraging inherent image prior information and are gaining traction in the field of medical image analysis. However, their application to feature representations of the non-salient objects, such as microvasculature, accessory organs, and early-stage tumors-is fundamentally limited by dimensional collapse problem, which diminishes feature diversity critical for non-salient structure discrimination. To address this, we propose a Multi-Granularity Masked Autoencoder (MG-MAE) framework for feature diversity learning: (1) We extend the conventional MAE into a multi-granularity framework, a global branch reconstructs global pixels, with a local branch recovering Histogram of Oriented Gradients (HOG) features, enabling hierarchical representation of both coarse-grained and fine-grained patterns; (2) Critically, in the local branch, a diversity-enhanced loss function incorporating Nuclear Norm Maximization (NNM) constraint to explicitly mitigate feature space collapse through orthogonal embedding regularization; and (3) A Dynamic Weight Adjustment (DWA) strategy that dynamically prioritizes hard-to-reconstruct regions via entropy-driven gradient modulation. Comprehensive evaluations across five clinical benchmarks-CCTA139, BTCV, LiTS, ACDC, and MSD Pancreas Tumour datasets-demonstrate that MG-MAE achieves statistically significant improvements in Dice Similarity Coefficient (DSC) scores for non-salient object segmentation, outperforming state-of-the-art methods. The code is available at https://github.com/zhangbbin/mgmae.
9 May 2026
Read appraisal →Canadian Association of Radiologists journal = Journal l'Association canadienne des radiologistes
Training With Local Data Remains Important for Deep Learning MRI Prostate Cancer Detection
OBJECTIVES: Domain shift has been shown to have a major detrimental effect on AI model performance however prior studies on domain shift for MRI prostate cancer segmentation have been limited to small, or heterogenous cohorts. Our objective was to assess whether prostate cancer segmentation models trained on local MRI data continue to outperform those trained on external data with cohorts exceeding 1000. METHODS: We simulated a multi-institutional consortium using the public PICAI dataset (PICAI-TRAIN: 1241 exams, PICAI-TEST: 259) and a local dataset (LOCAL-TRAIN: 1400 exams, LOCAL-TEST: 308). IRB approval was obtained and consent waived. We compared nnUNet-v2 models trained on the combined data (CENTRAL-TRAIN) and separately on PICAI-TRAIN and LOCAL-TRAIN. Accuracy was evaluated using the open source PICAI Score on LOCAL-TEST. Significance was tested using bootstrapping. RESULTS: Just 22% (309/1400) of LOCAL-TRAIN exams would be sufficient to match the performance of a model trained on PICAI-TRAIN. The CENTRAL-TRAIN performance was similar to LOCAL-TRAIN performance, with PICAI Scores [95% CI] of 65 [58-71] and 66 [60-72], respectively. Both of these models exceeded the model trained on PICAI-TRAIN alone which had a score of 58 [51-64] (P < .002). Reducing training set size did not alter these relative trends. CONCLUSION: Domain shift limits MRI prostate cancer segmentation performance even when training with over 1000 exams from 3 external institutions. Use of local data is paramount at these scales.
9 May 2026
Read appraisal →Abdominal radiology (New York)
Advancing osteoporosis opportunistic screening: multicenter validation of a deep learning algorithm using abdominal CT scans
OBJECTIVES: To develop and do multicenter validation on an algorithm that screens for osteoporosis from abdominal CTs. METHODS: This is a diagnostic accuracy study with retrospective data from January 2022 to July 2022 consisting of two steps: a segmentation step of the lumbar vertebral bodies, involving outpatient non-contrast abdominal CTs from Diagnósticos da América S/A (Dasa), and a multicenter validation step incorporating data from four additional institutions. The segmentation employed a 2D UNet with a ResNet34 backbone. We determined the Pearson correlation coefficient (r) between the mean of the slices' mean attenuations (MSMA) on CT scans against the bone mineral density (BMD) on recent DEXA scans, calculated AUCs and performance metrics for osteoporosis prediction, including 95% confidence intervals, and evaluated calibration plots. RESULTS: The multicenter validation included 504 participants (median age, 66 years, interquartile range 56-72; 388 women). A linear regression analysis showed an r of 0.62 (95% CI, 0.56-0.67) between MSMA (HU) and BMD (g/cm2). The AUCs (95% CI) for distinguishing between normal and osteoporosis were 0.96 (0.89, 1.0) for the internal dataset and 0.82 (0.74, 0.89) for the external dataset, and the performance metrics (95% CI), for a globally optimized threshold of 202.6 HU, were 100% sensitivity (44, 100) and 91% specificity (84, 95) for internal data and 79% sensitivity (61, 90) and 81% specificity (76, 84) for external sites. MSMA was independently associated with osteoporosis on a mixed-effects logistic regression analysis. The model showed good calibration, with Brier score of 0.054. CONCLUSION: We developed and performed a multicenter validation of a DL model for osteoporosis prediction on CT.
6 May 2026
Read appraisal →IEEE transactions on bio-medical engineering
Multilevel Correlation-Aware and Modal-Aware Graph Convolutional Network for Diagnosing Neurodevelopmental Disorders
OBJECTIVE: Graph-based methods using resting-state functional magnetic resonance imaging demonstrate strong capabilities in modeling brain networks. However, existing graph-based methods often overlook inter-graph relationships, limiting their ability to capture the intrinsic features shared across individuals. Additionally, their simplistic integration strategies may fail to take full advantage of multimodal information. To address these challenges, this paper proposes a Multilevel Correlation-aware and Modal-aware Graph Convolutional Network (MCM-GCN) for the reliable diagnosis of neurodevelopmental disorders. METHODS: At the individual level, we design a correlation-driven feature generation module that incorporates a pooling layer with external graph attention to perceive inter-graph correlations, generating discriminative brain embeddings and identifying disease-related regions. At the population level, to deeply integrate multimodal and multi-atlas information, a multimodal-decoupled feature enhancement module learns unique and shared embeddings from brain graphs and phenotypic data and then fuses them adaptively with graph channel attention for reliable disease classification. RESULTS: Extensive experiments on two public datasets for Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactivity Disorder (ADHD) demonstrate that MCM-GCN outperforms other competing methods, with an accuracy of 93.11% for ASD and 76.41% for ADHD. CONCLUSION: The MCM-GCN framework integrates individual-level and population-level analyses, offering a comprehensive perspective for neurodevelopmental disorder diagnosis, significantly improving diagnostic accuracy while identifying key indicators. SIGNIFICANCE: These findings highlight the potential of the MCM-GCN for imaging-assisted diagnosis of neurodevelopmental diseases, advancing interpretable deep learning in medical imaging analysis.
6 May 2026
Read appraisal →Magnetic resonance in medicine
Accelerating Multiparametric Quantitative MRI Using Self-Supervised Scan-Specific Implicit Neural Representation With Model Reinforcement
PURPOSE: To develop a self-supervised scan-specific deep learning framework for reconstructing accelerated multiparametric quantitative MRI (qMRI). METHODS: We propose REFINE-MORE (REference-Free Implicit NEural representation with MOdel REinforcement), combining an implicit neural representation (INR) architecture with a model reinforcement module that incorporates MR physics constraints. The INR component enables informative learning of spatiotemporal correlations to initialize multiparametric quantitative maps, which are then further refined through an unrolled optimization scheme enforcing data consistency. To improve computational efficiency, REFINE-MORE integrates a low-rank adaptation strategy that promotes rapid model convergence. We evaluated REFINE-MORE on accelerated multiparametric quantitative magnetization transfer imaging for simultaneous estimation of free water spin-lattice relaxation, tissue macromolecular proton fraction, and magnetization exchange rate, using both phantom and in vivo brain data. RESULTS: Under 4× and 5× accelerations on in vivo data, REFINE-MORE achieved superior reconstruction quality, demonstrating the lowest normalized root-mean-square error and highest structural similarity index compared to baseline methods and other state-of-the-art model-based and deep learning approaches. Phantom experiments further showed strong agreement with reference values, underscoring the robustness and generalizability of the proposed framework. Additionally, the model adaptation strategy improved reconstruction efficiency by approximately fivefold. CONCLUSION: REFINE-MORE enables accurate and efficient scan-specific multiparametric qMRI reconstruction, providing a flexible solution for high-dimensional, accelerated qMRI applications.
6 May 2026
Read appraisal →Annals of surgical oncology
CT-based Prediction of Visceral Pleural Invasion in Lung Adenocarcinoma ≤ 3 cm: Enhancing Deep Learning Specificity by Waiving Chest Wall Information
BACKGROUND: Accurate preoperative prediction of visceral pleural invasion (VPI) in lung adenocarcinoma is essential for guiding surgical decision-making. However, existing prediction models often sacrifice specificity when optimized for high sensitivity, increasing the risk of overtreatment. This study aimed to develop a computed tomography (CT)-based deep learning (DL) model that improves specificity by excluding chest wall information. METHODS: We retrospectively enrolled 835 patients with pathologically confirmed lung adenocarcinoma who underwent complete resection at two medical centers. A VPI-DL model was developed using a four-layer convolutional neural network incorporating a novel attention mechanism trained on chest-wall-masked CT inputs. The model was trained on 692 cases and externally validated on an independent cohort of 143 patients. Performance was benchmarked against the consolidation-to-tumor ratio (CTR), a published DL model, and assessments by three thoracic surgeons. Evaluation metrics included area under the curve (AUC), sensitivity, specificity, and accuracy under a high-sensitivity operating threshold. RESULTS: In external validation, VPI-DL achieved an AUC of 0.91, sensitivity of 91%, specificity of 82%, and accuracy of 83%, outperforming the CTR (specificity 63%), the previously published model (specificity 54%), and thoracic surgeons (specificity 57%) at comparable sensitivity levels. The attention-guided architecture effectively reduced spurious correlations and improved interpretability. CONCLUSION: The proposed VPI-DL model improves the specificity of VPI prediction while preserving high sensitivity. By intentionally excluding chest wall information, the model offers a robust and interpretable tool to aid preoperative planning and minimize the risk of both under- and overtreatment in early-stage lung adenocarcinoma.
6 May 2026
Read appraisal →European radiology
Enhancing vertebral fracture prediction using multitask deep learning computed tomography imaging of bone and muscle
OBJECTIVES: To develop and externally validate a computed tomography (CT)-based multitask learning model to predict fracture risk. MATERIALS AND METHODS: This study was conducted in two parts, using a multitasking learning approach. We developed a cross-sectional vertebral fracture (VF) detection model using abdominal CT scans of 2553 patients aged 50-80 years. Then, we leveraged this detection model within a multitask learning framework to develop a longitudinal VF prediction model over a 5-year follow-up period. External testing was performed on 1506 patients from two independent hospitals. The performance was compared between the single-task and multitask models, bone-only and bone+muscle images, and image-only and clinical models. RESULTS: For the cross-sectional fracture detection model, the mean age of the patients was 76.2 years, and 66.7% were female. In the classification task for detection of VF, the model using both bone and muscle showed an area under the receiver operating characteristic curve (AUROC) of 0.82 in the development set and 0.80 in the external test sets. Using multitask learning, the bone + muscle image model showed a c-index of 0.68 and had superior performance than the bone-only model in the external test set for 2-year, 3-year, and 5-year AUROCs (0.79 vs. 0.75, 0.71 vs. 0.68, and 0.71 vs. 0.68, respectively, all p < 0.01). Also, the multitask model significantly outperformed the Fracture Risk Assessment Tool (FRAX) (c-index: 0.68 vs. 0.66, p < 0.01). CONCLUSION: The CT-based multitask learning model integrating both bone and muscle data showed superior predictive performance for VFs compared with models using bone images only and traditional clinical models. KEY POINTS: Question Vertebral fracture risk remains underestimated in many individuals undergoing CT scans for other reasons, highlighting the need for improved opportunistic prediction tools. Findings A multitask deep learning model integrating both bone and muscle features from CT scans demonstrated superior performance compared to bone-only and traditional clinical models, including FRAX. Clinical relevance The proposed model enables accurate vertebral fracture risk prediction using routinely acquired CT scans, facilitating early identification and intervention without the need for additional tests.
6 May 2026
Read appraisal →Radiology. Imaging cancer
ONCO-RADS-guided Large Language Models for Extraction and Classification of Incidental Findings on Whole-Body Imaging Reports
Purpose To evaluate large language model (LLM)-based strategy performance for extraction and classification of incidental findings from whole-body (WB) imaging reports, particularly strategies incorporating Oncologically Relevant Findings Reporting and Data System (ONCO-RADS). Materials and Methods In this retrospective bicenter study, authors included all WB MRI reports from January 2016 to December 2023 at a referral center (internal dataset). Two observers extracted all incidental findings, and patient records were used to confirm final diagnoses. First, authors evaluated ONCO-RADS performance and the reproducibility of its incidental finding classifications by six radiologists. Then, authors evaluated the accuracy of three LLM-based strategies: (a) a fine-tuned DeBERTa/medical named entity recognition (NER) model; (b) zero-shot LLMs (ChatGPT-o1 [OpenAI], Gemini-2.5-Pro [Google]); and (c) reference-guided prompting of these LLMs using ONCO-RADS. Authors then expanded these strategies to an external dataset of 605 reports with multiple imaging techniques (405 WB MRI; 100 fluorodeoxyglucose PET/CT; and 100 chest-abdomen-pelvis CT acquisitions) from January 2022 to January 2025. Results The internal dataset included 823 patients (mean age, 63.7 years ± 11.7 [SD]; 457 male patients) with 1488 WB MRI reports. The average interobserver reproducibility of ONCO-RADS incidental finding classifications was excellent (Cohen κ, 0.87). The per-report accuracies of ONCO-RADS-guided LLMs (95.6% [151 of 158] and 86.7% [137 of 158] for ChatGPT-o1 and Gemini-2.5-Pro, respectively) were higher than those of the medical NER (69.0% [109 of 158]) and zero-shot LLMs (57.0% [90 of 158] and 70.9% [112 of 158] for ChatGPT-o1 and Gemini-2.5-Pro, respectively) (P < .001). In the external test set (mean age, 60.6 years ± 12.9; 330 male patients), the per-report accuracies of ONCO-RADS-guided ChatGPT-o1 (83.5% [505 of 605]) and Gemini-2.5-Pro (82.0% [496 of 605]) were higher than those of the models without ONCO-RADS prompting (63.1% [382 of 605] and 61.2% [370 of 605], respectively) and the medical NER (55.7% [337 of 605]) (P < .001). Conclusion Reference-guided prompting of the LLMs ChatGPT-o1 and Gemini-2.5-Pro with ONCO-RADS improved their performance in extracting and classifying incidental findings on WB imaging reports compared with zero-shot prompting and medical NER. Keywords: Large Language Models, Incidental Findings, Whole-Body MRI Supplemental material is available for this article. © RSNA, 2026.
6 May 2026
Read appraisal →Artificial intelligence in medicine
TADynFed: Dynamic modality-adaptive federated learning with tissue-aware disentanglement for cross-disease analysis
Federated learning (FL) enables collaborative medical image analysis across decentralized institutions while preserving data privacy. However, real-world deployment faces critical challenges: modality heterogeneity, where clients possess incomplete or varying medical image modalities, and cross-disease generalization, requiring models to adapt to unseen pathologies with anatomical consistency. Existing methods like FedAvg, FedProx, IOP-FL, and PntTranForFL assume uniform modality availability and static client participation, leading to poor performance under realistic clinical constraints. We propose TADynFed, a novel framework for the Heterogeneous Federated Learning (HFL) paradigm, which addresses both data and modality heterogeneity through, a tissue-aware disentanglement strategy that decouples modality-tailored and modality-shared features, a dynamic prototype memory bank for missing modality compensation, an adaptive aggregation mechanism that accounts for client reliability and tenure. We evaluate TADynFed using multi-disease MRI datasets including BraTS21 along with cross-domain imaging data from CheXpert (chest X-ray) and Hep-2 (microscopy), simulating a 13-client FL environment. The proposed framework achieves an average mDice score of 66.03%, significantly outperforming baseline methods such as PointTransformerFL at 58.60% and FedAvg at 53.06%. It also attains the lowest boundary alignment error with an ASD of 1.85 mm and HD95 of 8.70 mm, surpassing existing approaches by notable margins. In terms of calibration stability, TADynFed records an ECE of 0.09, indicating superior confidence reliability. Furthermore, it demonstrates high communication efficiency with only 76 MB of data exchanged per round, compared to 95-105 MB in other frameworks. These results validate TADynFed's ability to maintain high segmentation accuracy, boundary precision, and calibration stability while minimizing bandwidth usage. Outperforming existing frameworks by significant margins, TADynFed demonstrates robust boundary alignment, superior calibration, and efficient communication. It also exhibits strong cross-disease transferability without retraining. By integrating structured representation decomposition, prototype-guided fusion, and client-adaptive learning, TADynFed establishes a new benchmark in realistic, heterogeneous, and mix-modal federated medical imaging systems. Code supporting this study TADynFed.
5 May 2026
Read appraisal →Journal of nuclear medicine : official publication, Society of Nuclear Medicine
Validation of a Deep-Learning Coregistration Framework for Long-Axial-Field-of-View PET/CT Using Low-Radiation-Exposure Protocols Across Various Tracers
The advent of long-axial-field-of-view (LAFOV) PET/CT systems has significantly improved whole-body imaging by providing higher sensitivity and extended torso coverage. However, PET/CT potential misalignment remains a challenge, potentially introducing artifacts and quantification errors. Moreover, PET protocols with reduced scanning duration and dose, as well as the use of ultra-low-dose CT (ULD-CT), are increasingly relevant in clinical practice and research. This study aimed to evaluate the robustness and generalizability of a deep-learning coregistration framework for PET/CT alignment using low-dose PET and ULD-CT across various tracers with a LAFOV system. Methods: In total, 63 scans with 4 different tracers (89Zr-trastuzumab, 15O-H2O, 18F-MC225, and 18F-FDG) were included to assess PET and CT alignment improvements. Further evaluation was performed by comparing CT scans coregistered to the original PET images (rCT) with those coregistered to low-count (50%, 25%, and 12.5% of original counts) PET images (LC-rCT). Dice similarity coefficient, Jaccard similarity coefficient, Hausdorff distance, and average surface distance were used as evaluation metrics. Furthermore, ULD-CT coregistered to PET (rULD-CT) was compared with low-dose CT coregistered to the same PET (rLD-CT) using the same metrics. PET accuracy was evaluated by calculating SUVs. Results: The robustness of the deep-learning coregistration framework was demonstrated in reduced PET counts and ULD-CT scenarios across tracers. All metrics indicated robust performance when comparing LC-rCT with rCT and rULD-CT with rLD-CT. Consistent SUVs across varying PET counts and ULD-CT conditions further validated the quantitative accuracy of the approach. Conclusion: This work highlights that neither the use of low-dose PET protocols nor ULD-CT compromise the performance of this coregistration framework across 4 tracers. For LAFOV PET, these findings support the feasibility of such a framework in imaging protocols with reduced radiation exposure.
5 May 2026
Read appraisal →The British journal of radiology
Automatic segmentation of clinical target volume for radiation therapy in breast-conserving patients and exploration of clinical factors influential to its performance
OBJECTIVES: To develop and validate a deep learning model for whole breast clinical target volume (CTV) contouring and evaluate clinical features affecting its performance. METHODS: Five datasets with 857 patients from a single center were used. Dataset 1 (n = 300) trained and tested the model. Dataset 2 (n = 10) evaluated contouring time and dosimetric parameters. Datasets 3 (n = 20) and 4 (n = 10) were for clinical evaluation. Dataset 5 (n = 517) identified clinical factors influencing auto-contouring accuracy. Model performance was assessed using Dice Similarity Coefficient (DSC) and 95th percentile Hausdorff Distance (HD95). RESULTS: The median DSC and HD95 for left- and right-sided models in Dataset 1 were 0.941, 1.75 mm and 0.937, 2.47 mm, respectively. In Dataset 2, both auto-contouring and auto-contouring with manual corrections were significantly faster than manual contouring (P = .005 for both), while still achieving clinically acceptable dosimetric results. In Dataset 3, two physicians rated automatic and manual contours as equivalent (P = .214, P = .075), while the other rated auto-contouring higher (P < .001). In Dataset 4, the auto-contouring model outperformed 1/5 physicians by DSC (P = .009) and 3/5 by HD95 (P = .015, P = .007, P = .017). In Dataset 5, peripheral tumor-bed and low-density breast tissue were associated with lower DSC (P < .001 for both) and higher HD95 (P < .001 for both). Cases without unfavorable factors performed better than those with (P < .001 for both). CONCLUSIONS: The proposed model demonstrated acceptable accuracy, consistency, and efficiency in breast CTV contouring. Peripheral tumor-bed and low-density breast tissue reduced auto-contouring performance. ADVANCES IN KNOWLEDGE: The characteristics of challenging cases in whole breast CTV auto-contouring should be identified.
5 May 2026
Read appraisal →Cancer medicine
Predicting Prognosis for Gastric Cancer Patients Receiving Neoadjuvant Treatment With Body Composition-Based Deep Learning
BACKGROUND: This study sought to develop an innovative body composition (BC)-based deep learning (DL) model to precisely evaluate survival in gastric cancer (GC) patients undergoing neoadjuvant treatment (NT). MATERIALS AND METHODS: This retrospective study included GC patients undergoing NT from two centers. CT images both pre-NT and post-NT were preprocessed, focusing on the automatic segmentation of subcutaneous fat, visceral fat, and skeletal muscle regions using TotalSegmentator. Delta Radiomics features were extracted using Pyradiomics. After feature fusion and selection, the optimal model is Naive Bayes (Rad model). A hybrid DL model was developed by combining ResNet18 and Transformer networks for feature extraction. The Clinic_Rad_DL model was constructed by combining clinical features, radiomic signatures, and DL signatures. The ExtraTree classifier was used for the Clinic_Rad_DL model, while a separate Cox regression model was developed for survival analysis using the same features. RESULTS: A total of 356 patients (mean age, 59 ± 10 years; 264 males [74.2%]) were enrolled and divided into training, validation, and test sets in a 7:2:1 ratio. The DL model outperformed the Rad model. The Clinic_Rad_DL model outperformed both Rad model and DL model, with AUC of 0.915, 0.890, and 0.890 in training, validation, and test sets, respectively. The Cox proportional hazards model showed C-index of 0.806, 0.803, and 0.819, effectively stratifying patients into high- and low-risk groups with significant survival differences. CONCLUSION: The study developed and validated a BC-based DL model to predict survival in GC patients undergoing NT, offering potential for personalized treatment strategies in clinical practice.
5 May 2026
Read appraisal →Menopause (New York, N.Y.)
Deciphering the predictors of endometrial nonbenign lesions in asymptomatic postmenopausal women via explainable machine learning
OBJECTIVES: Timely identification of endometrial nonbenign lesions led to improved outcomes, but there was a lack of effective predictive models for asymptomatic endometrial thickening. The aim of this study was to develop a strong machine learning (ML) model for assessing the risk of endometrial malignancy in asymptomatic patients after menopause. METHODS: This retrospective study was designed to collect data from 971 postmenopausal asymptomatic women with endometrial thickening. The bootstrap resampling method was used for model training, internal validation, and external validation. With 41 easily accessible characteristics, multifactor regression and least absolute shrinkage and selection operator regression were performed for feature selection. Nine ML algorithms were applied to build a model. To explain the final model and rank feature importance, the SHapley Additive exPlanation (SHAP) method was utilized. Meanwhile, a nomogram was developed to facilitate model interpretation. RESULTS: The comprehensive methodologies identified parity, Doppler flow signals, endometrial thickness, cancer antigen 125, and D-dimer as significant predictors. The logistic regression (LR) model demonstrated superior performance compared with other ML algorithms, achieving an accuracy of 88%, a sensitivity of 78%, a specificity of 98%, and an area under the receiver operating characteristic curve of 0.81. Furthermore, individualized predictions of endometrial malignancy were visualized through a force plot generated by SHAP analysis. A nomogram based on the LR model was subsequently constructed, showing area under the receiver operating characteristic curve values of 0.82, 0.82, and 0.81 for the training, internal validation, and external validation cohorts, respectively. The calibration curve demonstrated excellent consistency. CONCLUSIONS: We developed an LR-based nomogram model and interpreted using the SHAP method, which provided visual insights for detecting endometrial nonbenign lesions in asymptomatic postmenopausal women. This approach would aid clinicians in providing individualized treatment and help avoid unnecessary invasive surgeries.
5 May 2026
Read appraisal →Pediatric radiology
Deep learning based gestational age estimation from multi-view fetal brain magnetic resonance imaging
BACKGROUND: Gestational age (GA) is essential for assessing fetal development, but conventional methods such as last menstrual period and ultrasound are often inaccurate, particularly in late pregnancy. Recent advances in deep learning (DL) and MRI offer more reliable and consistent GA estimation by capturing detailed fetal brain development. OBJECTIVE: This study aimed to develop deep learning models for GA prediction using multi-view fetal brain MRI and to compare their performance with conventional biometric regression techniques. MATERIALS AND METHODS: A total of 1,321 fetal MRI scans were used to train and evaluate various DL models, while an additional 80 publicly available MRI scans served as an external test set. Two training strategies were explored: transfer learning versus training from scratch, and single-view versus multi-modality input. RESULTS: The pre-trained ResNet-101 model achieved a mean absolute error (MAE) of 4.47 days and a coefficient of determination (R2) of 0.96 on the internal test set. On the external test set, the model yielded an MAE of 6.57 days, outperforming the biometric regression method, which achieved an MAE of 9.42 days. Explainability analysis revealed that the model predominantly focused on the lateral ventricles, cerebellum, and surrounding brain regions for GA prediction. CONCLUSIONS: The integration of multi-view MRI and transfer learning significantly enhanced the predictive accuracy of DL models for GA estimation. The proposed approach outperformed conventional biometric regression and highlighted clinically relevant anatomical regions, demonstrating its potential for use in prenatal diagnostic applications.
5 May 2026
Read appraisal →Abdominal radiology (New York)
A CT-based 2.5D deep learning model for preoperative T-staging in gastric cancer: a retrospective multicenter study
PURPOSE: To develop and validate a 2.5D multi-angle deep learning (MADL) model for preoperative T-staging in patients with gastric cancer (GC) and to explore the predictive potential for survival outcome. MATERIALS AND METHODS: In this retrospective study, GC datasets from four centers were used to develop a quaternary for preoperative T-staging model by integrating CT images with nine 2D slicers and three angles (0°, + 45°, -45°) from transverse, sagittal, and coronal views. Diagnostic performance (accuracy, sensitivity, specificity, F1-score) were compared between the 2.5D MADL and radiologists. The prognostic potential of 2.5D MADL-derived features was evaluated using Kaplan-Meier analysis and multivariate Cox regression. RESULTS: A total of 433 patients were divided into the training (n = 346, mean age, 64.8 years ± 10.5 [SD]; 95 female) and the internal validation (n = 87, mean age, 63.0 years ± 11.9; 23 female) sets from Centers 1-3, and 41 patients (mean age, 64.6 years ± 8.9; 10 female) formed the external testing set from Center 4. The 2.5D MADL model exhibited excellent diagnostic performance for T-staging in external testing set (AUCs of T1-T4: 0.962, 0.722, 0.913, 0.962). For T1 staging, it outperformed radiologists in accuracy (95% vs. 83%), sensitivity (75% vs. 12%), and F1 score (86% vs. 22%) in external testing set. Survival analysis showed significant overall survival differences between high- and low-risk groups stratified (P < 0.001), with high-risk patients having shorter survival (HR = 1.632, 95% CI:1.239-2.150). CONCLUSION: The 2.5D MADL model achieves accurate and reliable preoperative T staging of gastric cancer, with performance comparable to radiologists and superior accuracy for early-stage disease.
5 May 2026
Read appraisal →Joint diseases and related surgery
High diagnostic accuracy of a resnet50-based deep learning model for osteochondral lesions of the talus on magnetic resonance imaging
OBJECTIVES: This study aims to evaluate the diagnostic performance of a ResNet50-based convolutional neural network (CNN) in detecting osteochondral lesions of the talus (OLTs) on magnetic resonance imaging (MRI) and to compare its efficacy between T1- and T2- weighted sequences. MATERIALS AND METHODS: A total of 219 ankle MRI scans were reviewed retrospectively, including 60 with confirmed OLTs and 159 without lesions. From each study, coronal and sagittal T1- and T2-weighted images were extracted and standardized to 224 × 224 pixels. Augmentation techniques were applied to strengthen model training. Data were divided into training, validation, and test sets in a 60:20:20 split. A ResNet50 model initialized with ImageNet weights was fine-tuned using crossentropy loss with class weighting. Diagnostic performance was summarized with accuracy, precision, recall, and F1-scores. RESULTS: The model performed better on T1 sequences, achieving an accuracy of 94.1% (95% confidence interval [CI] 88.3-97.1%) and an area under the curve [AUC] of 0.93 (95% CI 0.87-0.97), with patient cases classified at 0.92 precision and 0.82 recall. Healthy controls in the T1 group were recognized with 0.95 precision and 0.98 recall. In contrast, T2 sequences were less reliable, showing an accuracy of 87.2% (95% CI 80.5-91.9%) and an AUC of 0.91 (95% CI 0.85-0.95). Precision for patient cases in the T2 group was notably lower (0.65) despite a recall of 0.81. Misclassifications were more frequent in the T2 dataset, as evidenced by the confusion matrices. CONCLUSION: Even with a relatively modest dataset, the ResNet50 model delivered strong results for T1-weighted MRI. While T2 images proved more challenging, suggesting that deep learning can add value to routine assessment of OLTs.
5 May 2026
Read appraisal →Journal of clinical ultrasound : JCU
Illuminating Research Dynamics: Medical Ultrasound and Deep Learning
BACKGROUND: Deep learning (DL) has enabled advances in ultrasound imaging, but challenges like limited datasets and device variability hinder progress. This study provides the first bibliometric overview of DL research in medical ultrasound. METHOD: We retrieved related publications (2004-Apr 2025) on medical ultrasound and DL from the Web of Science Core Collection. Bibliometric tools (Bibliometrix, VOSviewer, CiteSpace) were used to quantify publication growth, country/institution contributions, and thematic evolution. RESULTS: A total of 3386 publications were included, with the annual output increasing steadily at an average growth rate of 31.02%. While China led all countries in publication volume (1355), the United Kingdom achieved the highest citation rate per article (55.5 citations). Fudan University was the most prolific institution (110), and key authors such as Luca Saba and Jasjit S. Suri formed the principal collaboration hubs. In terms of venues, Diagnostics published the most papers, whereas IEEE Transactions on Medical Imaging exerted the greatest citation impact (5646 citations; IF 8.9). Keyword analysis revealed core themes in classification, segmentation, and CNN-based diagnosis. From 2020 onward, ultrasound-specific topics (e.g., diagnosis, images) dominate; advanced architectures (e.g., U-Net) and emerging clinical concepts such as noninvasive assessment. CONCLUSION: This mapping highlights a rapidly growing, internationally collaborative field. Emerging trends include advanced artificial intelligence (AI) architectures (U-Net, attention-based networks, vision Transformers) and a shift toward surgical and therapeutic applications. These insights underscore the need for multicenter validation, standardized data protocols, and clinically interpretable AI models to accelerate safe translation into practice.
5 May 2026
Read appraisal →Magnetic resonance in medicine
Dynamic-Guided Diffusion Probability Model for Cranial Nerves Segmentation
PURPOSE: Segmentation of cranial nerves (CNs) bundles using magnetic resonance imaging (MRI) provides a valuable quantitative approach for analyzing the morphology and orientation of individual CNs. Currently, the CN regions can be segmented directly using deep learning-based methods. However, existing methods overlook the unique characteristics of CNs, particularly their environmental features and representation in multimodal images that may lead to suboptimal segmentation outcomes. METHODS: We proposed a dynamic-guided diffusion probability model for CNs segmentation, which enhances segmentation performance by integrating the intrinsic characteristics of CNs. A dynamic-guided mechanism approach called the SE-A-NL module was proposed. The module is capable of addressing both the varying characterization abilities of multimodal images and the long-range connections of CNs within images. RESULTS: Quantitative and qualitative experiments demonstrate that the proposed method surpasses current state-of-the-art approaches, delivering accurate and effective segmentation of five pairs of cranial nerves. Notably, the method outperforms existing techniques in 16 out of the 20 evaluated metrics. CONCLUSION: The overall network model effectively integrates multimodal information and anatomical priors by combining multi-channel attention and non-local attention mechanisms, thereby improving CNs segmentation performance. Thorough comparative and ablation studies highlight the superior performance of the proposed method.
5 May 2026
Read appraisal →Abdominal radiology (New York)
Deep learning-based combined noise reduction and contrast enhancement for post-neoadjuvant pancreatic cancer CT: does improved image quality translate to better resectability assessment?
PURPOSE: To evaluate whether deep learning-based combined noise reduction and contrast enhancement reconstruction (DLR) improves image quality and resectability prediction accuracy compared to conventional iterative reconstruction (IR) in post-neoadjuvant pancreatic cancer CT assessment. METHODS: This retrospective study included 114 patients with pancreatic cancer following neoadjuvant therapy. Contrast-enhanced CT images were reconstructed using conventional IR and vendor-neutral ClariACE. Three abdominal radiologists independently assessed image quality (based on 8 parameters: tumor conspicuity, tumor margin, image noise, sharpness of the main pancreatic duct, arterial depiction, venous depiction, plasticity and overall image quality) and determined tumor resectability with confidence levels. Quantitative analysis included aortic and portal venous attenuation measurements and pancreas-to-tumor contrast-to-noise ratio (CNR). Diagnostic performance was evaluated using the area under the receiver operating characteristic curve (AUC) with DeLong's test. Sensitivity, specificity, accuracy, and reader confidence were compared using McNemar's test. RESULTS: DLR demonstrated significantly superior vessel enhancement (p < 0.001) and improved CNR (p < 0.001) versus conventional IR. Two readers consistently rated DLR images higher across all qualitative categories (p < 0.001) except plasticity, while the third reader favored DLR in five of eight parameters (p < 0.001 to 0.020). However, all readers noted increased artificial appearance in DLR images (p < 0.001). Despite image quality improvements, no significant differences were observed in resectability assessment accuracy (62.3%-65.8%), AUC values (0.485-0.520), or high-confidence diagnosis rates between reconstruction methods. CONCLUSION: Although deep learning-based combined noise reduction and contrast enhancement reconstruction significantly improved quantitative and subjective image quality metrics, it did not enhance diagnostic accuracy for predicting R0 resectability in post-neoadjuvant pancreatic cancer patients.
5 May 2026
Read appraisal →Journal of vascular and interventional radiology : JVIR
A Deep Learning Model for Artery of Adamkiewicz Detection in Bronchial Artery Embolization: A Multicenter Retrospective Study
PURPOSE: To evaluate the effectiveness of a deep learning model for recognizing the artery of Adamkiewicz and anterior spinal artery (ASA) to prevent spinal artery non-target embolization during bronchial artery embolization (BAE). MATERIALS AND METHODS: This multicenter, retrospective study included 2036 patients from January 2019 to December 2023 with hemoptysis who underwent de novo BAE. A deep learning-based framework was proposed for spinal artery identification, comprising region of interest (ROI) perception and target spinal artery identification. The ROI perception extracted vessel-related regions to improve the conspicuity of the spinal artery. The target spinal artery identification utilized a progressive refinement learning network, localizing the artery from the global view, and progressively refined the identification results through cross-scale information interaction. RESULTS: Seventy-eight patients (3.8%) had an identifiable artery of Adamkiewicz and ASA on right intercostal-bronchial artery angiography. The sensitivity of the proposed method was 92.1% and the specificity was 84.6%, with no statistically significant difference from a fellow radiologist interpretation. CONCLUSIONS: The deep learning system for spinal artery detection during BAE exhibited high sensitivity and performance, comparable to radiology fellows.
4 May 2026
Read appraisal →European radiology
A deep learning model to reduce agent dose for contrast-enhanced MRI of the cerebellopontine angle cistern
OBJECTIVES: To evaluate a deep learning (DL) model for reducing the agent dose of contrast-enhanced T1-weighted MRI (T1ce) of the cerebellopontine angle (CPA) cistern. MATERIALS AND METHODS: In this multicenter retrospective study, T1 and T1ce of vestibular schwannoma (VS) patients were used to simulate low-dose T1ce with varying reductions of contrast agent dose. DL models were trained to restore standard-dose T1ce from the low-dose simulation. The image quality and segmentation performance of the DL-restored T1ce were evaluated. A head and neck radiologist was asked to rate DL-restored images in multiple aspects, including image quality and diagnostic characterization. RESULTS: 203 MRI studies from 72 VS patients (mean age, 58.51 ± 14.73, 39 men) were evaluated. As the input dose increased, the structural similarity index measure of the restored T1ce increased from 0.639 ± 0.113 to 0.993 ± 0.009, and the peak signal-to-noise ratio increased from 21.6 ± 3.73 dB to 41.4 ± 4.84 dB. At a 10% input dose, using DL-restored T1ce for segmentation improved the Dice from 0.673 to 0.734, the 95% Hausdorff distance from 2.38 mm to 2.07 mm, and the average surface distance from 1.00 mm to 0.59 mm. Both DL-restored T1ce from 10% and 30% input doses showed excellent image quality (3, interquartile range (IQR) [Q3-Q1] = 3-3 and 3, IQR [Q3-Q1] = 4-3), with the latter being considered more informative (4, IQR [Q3-Q1] = 4-3). CONCLUSION: The DL model improved the image quality of low-dose MRI of the CPA cistern, which makes lesion detection and diagnostic characterization possible with 10-30% of the standard dose. KEY POINTS: Question Deep learning models that aid in the reduction of contrast agent dose are not extensively evaluated for MRI of the cerebellopontine angle cistern. Findings Deep learning models restored the low-dose MRI of the cerebellopontine angle cistern, yielding images sufficient for vestibular schwannoma diagnosis and management. Clinical relevance Deep learning models make it possible to reduce the use of gadolinium-based contrast agents for contrast-enhanced MRI of the cerebellopontine angle cistern.
4 May 2026
Read appraisal →Journal of medical imaging and radiation sciences
Explainability-informed benchmarking of two deep learning models for organ-at-risk segmentation in MR-guided adaptive radiotherapy
INTRODUCTION/BACKGROUND: Segmentation of gastrointestinal (GI) organs-at-risk (OARs) is a critical yet time-consuming step in MR-guided adaptive radiotherapy (MRgRT), with manual delineation prone to inter- and intra-observer variability. While deep learning approaches have shown promise, their clinical adoption requires not only accuracy but also interpretability and reliability. This study benchmarks two widely used convolutional architectures, U-Net and Residual U-Net (ResUNet), for abdominal OAR segmentation, with an emphasis on explainability-oriented quantitative analysis. METHODS: An anonymized abdominal MRI dataset was used to train and evaluate U-Net and ResUNet using a 5-fold stratified group cross-validation strategy. Segmentation performance was assessed using the Dice Similarity Coefficient (DSC), Intersection-over-Union (IoU), and the 95th percentile Hausdorff Distance (HD95). Explainability was investigated using Gradient-weighted Class Activation Mapping (Grad-CAM) computed from the final convolutional layer of each network. To enable objective analysis beyond qualitative visualization, Grad-CAM activation maps were quantified using numerical localization metrics relative to ground-truth organ masks, including in-organ energy ratio, boundary energy ratio, pointing accuracy, activation Dice coefficient, centroid distance and activation entropy. Grad-CAM metrics were aggregated across gastrointestinal organs and averaged over the five validation folds. RESULTS: Both architectures demonstrated comparable segmentation performance across organs, with no statistically significant differences across evaluated metrics. Grad-CAM analysis showed similar region-level attention patterns, with in-organ activation ratios of 71.4 ± 8.6% for U-Net and 66.2 ± 9.1% for ResUNet, boundary energy ratios of 24.1 ± 4.9% and 21.8 ± 5.2%, respectively, and pointing accuracies exceeding 70% for both models. Uncertainty analysis based on inter-fold variability and boundary error dispersion indicated comparable stability and bounded worst-case behavior. DISCUSSION/CONCLUSION: By integrating performance, uncertainty and explainability quantitative indicators, this study provides an informed benchmarking of two deep learning models for abdominal OAR segmentation. The results suggest that both U-Net and ResUNet exhibit stable and interpretable behavior under the evaluated configurations, supporting their potential use in MR-guided adaptive radiotherapy workflows where reliability and clinical trust are essential.
4 May 2026
Read appraisal →Journal of medical imaging and radiation sciences
Automatic field-of-view planning for magnetic resonance shoulder imaging using Deep Learning
INTRODUCTION: Accurate prescription of oblique coronal and oblique sagittal field of views (FOV) is essential for diagnostic shoulder MRI. Manual planning is radiographer-dependent, time-consuming, and subject to inter- and intra-operator variability, leading to inconsistent image quality and incomplete coverage. Although deep learning (DL) has advanced automated scan planning in non-oblique planes, oblique shoulder prescriptions remain underexplored; an automated DL approach could standardize FOV prescription, reduce operator dependence, and improve reproducibility and workflow without compromising diagnostic quality. METHODS: In this retrospective multicenter study, 575 shoulder MRI examinations (2019-2025) from four sites were included. Sites A (n=151) and B (n=220) were used for training; testing was performed on sites C (n=61), and D (n=143). A two-stage pipeline was implemented using five oriented bounding box (OBB) variants of YOLOv11 (n, s, m, l, x): Stage 1 performed slice selection; Stage 2 performed FOV prescription. Performance was evaluated against radiographers' prescriptions using mean absolute slice difference (MASD, slices), intersection over union (IoU), and mean absolute angle difference (MAAD, degrees). Clinical utility was assessed by three raters. RESULTS: The YOLOv11-OBB-l model achieved the lowest MASD for Stage 1 (1.016±0.153 slices). For Stage 2, YOLOv11-OBB-x performed best (coronal IoU, 0.847±0.003; sagittal IoU, 0.852±0.007; MAAD, 3.259±0.190°). During testing across each site, MASD ranged from 0.700±0.837 to 1.192±2.550 slices; MAAD from 2.811±2.348 to 4.396±7.158°; coronal IoU from 0.800±0.092 to 0.872±0.065; and sagittal IoU from 0.824±0.111 to 0.887±0.047. Mean clinical utility was 97.2%. Performance was noninferior to interrater variability across all sites and metrics. CONCLUSION: DL-based automated FOV prescription for shoulder MRI achieves performance comparable to radiographers, generalizes across institutions, and demonstrates high clinical utility.
4 May 2026
Read appraisal →Neural networks : the official journal of the International Neural Network Society
MMC-CS: Multi-branch multi-stage contrastive learning for self-supervised compressed sensing
In recent years, optimization-inspired networks that integrate optimization theory into deep neural networks (DNNs) have achieved remarkable success in image compressed sensing (ICS). However, current DNN-based ICS methods face two key challenges: (1) the difficulty of acquiring ground-truth (GT) data from labeled measurements and (2) the underutilization of measurement data. To address these issues, a novel self-supervised deep learning framework is proposed to solve the inverse problem in ICS. Building upon traditional self-supervised learning, our proposed method leverages measurement values through a multi-branch, multi-stage progressive cross-contrast structure, enabling effective learning of the common prior of the original image in the absence of GT data. Based on this framework, a novel Multi-branch Multi-stage Cross-contrast CS (MMC-CS) end-to-end DNN is designed, which unfolds the Proximal Gradient Descent (PGD) algorithm. Our approach further implements multi-scale co-optimization in the image reconstruction process, integrating both the image paths (IPs) and convolutional feature paths (CFPs). Additionally, the reconstruction performance across multiple spatial levels is enhanced by incorporating wavelet convolution (WTConv). Extensive experiments demonstrate that our proposed method outperforms existing self-supervised approaches, achieving an average PSNR improvement of 0.3-1.6 dB. Moreover, it shows strong potential to compete with state-of-the-art supervised methods.
4 May 2026
Read appraisal →NMR in biomedicine
Deep Learning Reconstruction Enables Diagnostic-Quality 0.4T Knee and Spine MRI in One-Third of the Time
There has been a growing interest in low-field MRI due to its lower costs, enabling an increase in accessibility of MRI worldwide. Long scan times are currently needed to obtain high-quality images. The aim of our study was to accelerate low-field 0.4T musculoskeletal MRI using deep learning and evaluate maximum acceleration via assessment by two expert radiologists. A cascades of independently recurrent inference machines (CIRIM) model was trained using knee and spine 2D multislice scans, retrospectively undersampled in k-space. The heterogeneous dataset included varied contrasts, orientations, and matrix sizes. To determine the maximum possible acceleration factor, the undersampling pattern and loss function of the CIRIM were optimized using the knee dataset. Models were trained for acceleration factors 1.5-4. The different acceleration factors were evaluated quantitatively and clinically, with radiologist scores compared using Kendall's Tau and Cohen's Kappa. To assess the model's generalizability, two prospectively undersampled scans were evaluated, and the knee-trained models were tested on spine data. CIRIM accurately reconstructed all data, with slight metric degradation at higher accelerations. Combining L1, structural similarity index measure (SSIM), and perceptual loss improved sharpness, while undersampling pattern differences were minimal. Clinical scores from both radiologists declined with acceleration, reflected in Kendall's Tau values of 0.64, 0.64, 0.54, and 0.55 for image quality, sharpness, anatomical conspicuity, and artifacts. Inter-observer agreement was low (Cohen's kappa: 0.227-0.299). The model generalized well, reconstructing spine data with a knee-trained model at comparable quality and successfully handling prospectively undersampled scans. CIRIM consistently delivered high-quality reconstructions, demonstrating excellent robustness to variations in contrast, orientation, matrix size, and anatomy. Its ability to generalize across anatomies and handle prospectively undersampled scans highlights its practical utility. The differences in reading for higher accelerations between radiologists suggest that the optimal acceleration factor is context-dependent and should be tailored to the specific diagnostic task.
4 May 2026
Read appraisal →Taiwanese journal of obstetrics & gynecology
Segmentation of uterus and myoma from MRI image based on deep learning
OBJECTIVE: Myoma is a common gynecologic condition with abnormal muscular and fibrotic tissue growth in the uterus. Compared with sonography, magnetic resonance imaging (MRI) provides more detailed information with a better diagnostic accuracy. However, the interpretation of the imagery is shrewdness- and experience-dependent. The therapeutic strategy of myoma can be directly affected by the relative location and size in the uterus. Hence, this study aimed to generate an objective standardized modality in interpreting the imagery. MATERIALS AND METHODS: Deep learning (DL) was used to train two models using MRI imagery labeled by radiologists: 1) Model Uterus + Endometrium: segmentation of uterus and endometrium; 2) Model Myoma: segmentation of myoma. The myomas were classified based on their relative position in the uterus according to FIGO criteria. Each set of MRI image sequence was obtained from uterine sagittal sections with 5.5-mm interval of 30 slices per case. The dataset was divided into training, validation, and test sets. The number of images was first increased by data augmentation before implementing a DL semantic segmentation DeepLabV3+model. Cross entropy and Dice function were used for the loss and evaluation functions, respectively. RESULTS: The Dice values of uterus, endometrium and myoma segmentation reached 81.55%, 73.49% and 67.34%, respectively. Thus, the 3-D prediction of the uterine, endometrial and myomatous regions had a positive effect on determining the classification of myomas. CONCLUSION: DL can be used as a tool to improve the diagnosis and classification of uterine myoma. This model can also be applied to other gynecologic diseases.
4 May 2026
Read appraisal →Annals of surgical oncology
Computed Tomography-Pathology Deep Learning Model for the Precise Prediction of Recurrence in Pathological Stage IA Lung Adenocarcinoma
BACKGROUND: The postoperative prognosis of pathological stage IA lung adenocarcinoma (LUAD) exhibits significant heterogeneity. While the tumor node metastasis (TNM) staging system offers limited recurrence prediction capability, this study aims to develop a precise deep learning model for prognostic stratification. PATIENTS AND METHODS: We retrospectively analyzed a consecutive cohort of patients with pathological stage IA LUAD who underwent surgical resection at Zhongshan Hospital Fudan University. Our novel ResNet 3D-Pathology Fusion (Res3D-PF) model-integrating a three-dimensional ResNet backbone with an image-pathology fusion module-was developed to predict recurrence-free survival (RFS) using preoperative computed tomography (CT) images and International Association for the Study of Lung Cancer (IASLC) grading. Model performance was evaluated through receiver operating characteristic curve analysis, with independent RFS predictors identified via multivariable Cox regression. RESULTS: Among 551 patients with stage IA LUAD (median age 61 years; 339 women) divided into training (n = 368) and validation (n = 183) sets, the CT-pathology fusion model achieved superior predictive performance. In the validation cohort, Res3D-PF significantly outperformed the 8th T-stage (AUC 0.837 vs. 0.660, p = 0.001) and IASLC grade (AUC 0.837 vs. 0.684, p = 0.015) for 5-year RFS prediction. Multivariable analysis confirmed Res3D-PF as an independent prognostic factor (HR 15.772, 95% CI 3.384-73.508; p < 0.001). Model-stratified high-risk patients demonstrated significantly reduced 5-year RFS (73.1% vs. low-risk 98.5%, p < 0.001) and shorter median RFS (74.4 vs. 96.2 months, p < 0.001). CONCLUSIONS: We developed and validated a CT-pathology deep learning model that outperforms conventional TNM staging and IASLC grading for predicting postoperative recurrence in stage IA LUAD. This approach enables individualized risk stratification to guide precision treatment strategies.
4 May 2026
Read appraisal →Journal of occupational and environmental medicine
Combining MEA-Net and LAP-Net for Pneumoconiosis Staging Framework
OBJECTIVE: Pneumoconiosis is a common and highly hazardous occupational disease. The staging of pneumoconiosis is mainly carried out by experienced doctors on the basis of the shadows and textures on lung x-ray films. Despite well-defined criteria, the process remains influenced by individual clinical judgment. METHODS: To improve the subjective process of pneumoconiosis diagnosis, this study proposes a new deep learning framework for pneumoconiosis staging framework using MEA-Net and LAP-Net. RESULTS: The experimental results show that the accuracy, precision, recall, specificity, F1-score, and area under the curve in the four-stage classification reached 95.24%, 95.15%, 95.15%, 90.58%, 94.85%, and 98.87%, respectively. CONCLUSION: The proposed method can help doctors to identify the different stages of pneumoconiosis more accurately in the diagnosis of the disease.
4 May 2026
Read appraisal →Radiology. Artificial intelligence
Radiopathomic Graph Deep Learning for Multiscale Spatial-Contextual Modeling of Intratumoral Heterogeneity to Predict Breast Cancer Response to Neoadjuvant Therapy
Purpose To develop an explainable radiopathomic graph deep learning (RPGDL) system for multiscale spatial-contextual modeling of intratumoral heterogeneity and evaluate its performance for the prediction of pathologic complete response (pCR) to neoadjuvant therapy in breast cancer. Materials and Methods The RPGDL system was developed from dual-center retrospective analysis of patients with biopsy-proven invasive breast cancer (May 2018-August 2024). For each tumor, individual radiomic and pathomic graphs were generated from pretherapeutic MRI and hematoxylin-eosin-stained biopsy slide images, respectively. These graphs were then processed by three distinct graph neural networks (GNNs): radiomic, pathomic, and radiopathomic. GNN performance was assessed with the area under the receiver operating characteristic curve (AUC), net reclassification index (NRI), and integrated discrimination improvement (IDI). A multifaceted approach was used to explain the GNNs' predictions. Results The training set included 582 (mean age, 52 years ± 9 [SD]) patients and the external test set 468 (50 years ± 10) patients from centers 1 and 2, respectively. The radiomic GNN achieved AUCs of 0.89 (95% CI: 0.85, 0.93) in the training set and 0.84 (95% CI: 0.80, 0.89) in the external test set; the pathomic GNN achieved AUCs of 0.87 (95% CI: 0.83, 0.91) in the training set and 0.83 (95% CI: 0.78, 0.88) in the external test set, with no significant difference between them (P > .05). The radiopathomic GNN outperformed both single-modality GNNs (training set: AUC, 0.95 [95% CI: 0.92, 0.98]; external test set: AUC, 0.91 [95% CI: 0.87, 0.94]; P < .05; NRI and IDI confirmed). Pathomic graphs dominated probability increases for pCR predictions, while radiomic graphs drove probability decreases for non-pCR predictions. Multifaceted analyses verified GNNs' explainability. Conclusion The developed RPGDL system enabled multiscale spatial-contextual intratumoral heterogeneity modeling for high-performance, explainable prediction of pCR to neoadjuvant therapy in breast cancer. Keywords: Dynamic Contrast-enhanced MRI, Breast, Tumor Response, Radiology-Pathology Integration, Prognosis, Principal Component Analysis, Perception, Supervised Learning, Reconstruction Algorithm Supplemental material is available for this article. © RSNA, 2026.
4 May 2026
Read appraisal →European radiology
Predicting molecular types of adult-type diffuse gliomas based on MRI reports with large language models
OBJECTIVES: To evaluate the performance of large language models (LLMs) in predicting molecular types of adult-type diffuse gliomas according to the 2021 WHO classification using MRI radiology reports. MATERIALS AND METHODS: This retrospective study included 2169 patients diagnosed with adult-type diffuse gliomas (294 oligodendrogliomas, 295 IDH-mutant astrocytomas, and 1580 IDH-wildtype glioblastomas) between July 2005 and March 2024 from four hospitals in Asia and Europe. Seven proprietary and open-source LLMs were assessed: GPT-4o-mini, GPT-4.1-mini, Llama 3.1 8B, Llama 3.1 70B, Qwen2.5 7B, Deepseek-r1 8B, and Mistal 7B. The performance of LLMs in classifying molecular types was compared based on the provision of relevant knowledge of glioma imaging findings (knowledge-based vs. naïve prompt). The impact of radiologists' subspecialization in neuro-oncology, report quality, and reporting language on LLMs' performance was also evaluated. RESULTS: LLMs achieved significantly higher (naïve vs. knowledge-based; GPT-4o-mini, 77.0% vs. 79.1%, p < 0.001; Qwen2.5 7B, 75.9% vs. 79.5%, p < 0.001; Deepseek-r1 8B, 66.0% vs. 73.2%, p < 0.001) or comparable accuracy (GPT-4.1-mini, 78.7% vs. 78.6%; Llama 3.1 70B, 78.0% vs. 78.1%; Mistral 7B, 58.4% vs. 57.4%) using knowledge-based prompt compared to naïve prompt, except for Llama 3.1 8B (65.4% vs. 44.6%, p < 0.001). Differences in accuracy were more pronounced in smaller-sized LLMs. Additionally, the accuracy was significantly higher with reports by neuro-oncology specialists and high-quality reports in all LLMs (p < 0.001). CONCLUSIONS: LLMs may provide preoperative information on the tumor types of adult-type diffuse gliomas from MRI reports by providing relevant knowledge in the prompt. Informative and descriptive reports could further enhance LLMs' performance. KEY POINTS: Question Our study aimed to evaluate large language models' (LLMs) ability to efficiently predict molecular types of adult-type diffuse gliomas according to the 2021 WHO classification. Findings Larger models generally showed better accuracy and were less sensitive to domain-specific knowledge. Their performance improved when using high-quality, longer reports or reports by neuro-oncology specialists. Clinical relevance These findings highlight the potential role of LLMs in predicting glioma molecular types, underscoring the importance of informative and descriptive reports in enhancing their performance.
4 May 2026
Read appraisal →Knee surgery, sports traumatology, arthroscopy : official journal of the ESSKA
Synthetic computed tomography from magnetic resonance imaging: An editorial on deep learning approaches for hip and knee image translation.
This editorial provides a concise overview of synthetic computed tomography (CT) generation from magnetic resonance imaging (MRI) for musculoskeletal care, with a focus on hip and knee applications. We outline why interest in this area is growing, including the potential to reduce radiation exposure, consolidate imaging into a single visit, and align measurements across modalities for preoperative planning and longitudinal follow-up. At a high level, we summarize how deep learning methods transform MRI into CT-like volumes and what that could mean for bone morphology, implant planning, and combined assessment of bone and soft tissue. We briefly introduce representative model classes, including conditional generative adversarial networks that pair realism objectives with voxelwise constraints and diffusion models that replace adversarial training with iterative denoising for stable optimization and strong anatomical fidelity. Our goal is to orient clinicians, engineers, and researchers to the key promises and open questions, and to motivate collaborative studies that test whether synthetic CT can deliver reliable, reproducible, and clinically useful information at the point of care.
4 May 2026
Read appraisal →PloS one
Interpretable multivariate survival models: Improving predictions for conversion from mild cognitive impairment to Alzheimer's disease via data fusion and machine learning
Accurately predicting which individuals with mild cognitive impairment (MCI) will progress to Alzheimer's disease (AD) can improve patient care. This study examines the role of quantitative MRI (qMRI), cognitive evaluations, apolipoprotein [Formula: see text]4 (APOE [Formula: see text]4), and cerebrospinal fluid (CSF) biomarkers in Cox survival models to predict progression from MCI to AD. Data from 564 participants in the ADNI study, who transitioned from MCI to AD, were analyzed. The data set included 330 features encompassing qMRI, cognitive assessments, CSF biomarkers, and APOE [Formula: see text]4 status. Advanced machine learning (ML) methods were applied to evaluate the importance of these data sources, select relevant features, and develop interpretable Cox survival models within a cross-validation framework. The top optimized model achieved a sensitivity of 0.69, 95% CI [0.63, 0.76], and a specificity of 0.87, 95% CI [0.83, 0.90], and used all data sources. The results demonstrated that combining qMRI features with cognitive assessments, CSF biomarkers, and APOE [Formula: see text]4 status, analyzed using the BSWiMS model, resulted in a substantial improvement in the ability to predict progression from MCI to AD, achieving 81% precision and 87% specificity. These results exceed those obtained with other models evaluated. Finally, biomarker analysis showed that cognitive scores are the most relevant features to predict conversion, followed by CSF and qMRI biomarkers. These findings highlight the value of integrating multiple data sources in highly interpretable Cox survival models for the early identification of individuals at risk for AD.
4 May 2026
Read appraisal →Magnetic resonance in medical sciences : MRMS : an official journal of Japan Society of Magnetic Resonance in Medicine
Visualization of the Trochlear Nerve Using Deep Learning-enhanced 3D T2-weighted MR Imaging at 3T
PURPOSE: Cranial nerve imaging with 3T MRI commonly uses 3D fast imaging employing steady-state acquisition (3D-FIESTA); however, this sequence has limitations in achieving higher spatial resolution and provides poor tissue contrast between cranial nerves and adjacent vascular structures. We evaluated 3D T2-weighted imaging (T2-CUBE) with deep learning-based reconstruction (DLR) for cranial nerve visualization, focusing on the trochlear nerve, the smallest cranial nerve with the longest intracranial course, and compared it with T2-CUBE without DLR and 3D-FIESTA. METHODS: Ten healthy male volunteers (age, 23-40 years; mean age, 32 years) underwent T2-CUBE with and without DLR, and 3D-FIESTA at 3T. Two neuroradiologists independently evaluated trochlear nerve visualization in 4 anatomical segments (origin from the midbrain, cisternal, tentorial, and anterior portion of its cavernous segments) using a 3-point scale, and SNR of the pons (SNRPONS) and cerebrospinal fluid (SNRCSF) were calculated. RESULTS: T2-CUBE with DLR achieved a 100% visualization across all trochlear nerve segments and demonstrated significantly better visualization than both T2-CUBE without DLR and 3D-FIESTA (P < 0.025). T2-CUBE without DLR showed 67.5%-100% visualization across the 4 segments. 3D-FIESTA showed 32.5%-80% visualization of the origin from the midbrain, cisternal, and tentorial segments, with no visualization of the cavernous segments. DLR increased SNRPONS and SNRCSF by factors of 1.8-2.5 (SNRPONS: 14.1 vs 5.7; SNRCSF: 31.8 vs 17.5, respectively; P < 0.001). T2-CUBE with DLR demonstrated significantly higher SNRPONS than 3D-FIESTA (14.1 vs 6.4, P < 0.001), while SNRCSF was comparable (31.8 vs 36.3, P = 0.20). CONCLUSION: T2-CUBE with DLR at 3T provided a significantly better trochlear nerve visualization than T2-CUBE without DLR and 3D-FIESTA. This technique may extend beyond the trochlear nerve to other cranial nerves and to the evaluation of neurovascular compression in the cistern, with the potential to become the new standard for cisternal imaging.
3 May 2026
Read appraisal →JMIR aging
Machine Learning-Derived Cardiovascular Aging Phenotypes From Cardiac Function and Stroke Risk in the UK Biobank: Cohort Study
BACKGROUND: Cardiovascular magnetic resonance (CMR) is widely used across various cardiac conditions and systematically assesses cardiac anatomical structures and functional dynamics. Machine learning (ML) can accurately predict outcomes and understand the inherent features of clinical data. OBJECTIVE: This study aimed to derive CMR phenotypes related to cardiovascular aging, investigate the relationship between these phenotypes and stroke risk, and relearn these phenotypes using supervised ML. METHODS: We enrolled 36,467 participants without stroke and extracted CMR parameters from the UK Biobank, with follow-up data extending until September 30, 2023. Using the generative topographic mapping technique, we identified latent grid nodes among participants and then derived phenotypes through agglomerative hierarchical clustering. We used supervised ML models to predict cardiac function phenotypes and used Cox proportional hazards models to assess the association between these phenotypes and long-term stroke risk. RESULTS: We enrolled 36,467 participants in the study. The mean age was 54.9 (SD 7.5) years, with 17,442 (47.8%) male participants. During a mean follow-up time of 14.7 (SD 1.1) years, 500 (1.4%) participants developed stroke and 664 (1.8%) participants died, respectively. After generative topographic mapping modeling, we identified 2 distinct phenotypes: phenotype 1, characterized by adverse cardiac function and an accumulation of cardiovascular risk factors, reflecting cardiovascular aging; and phenotype 2, associated with a lower risk of stroke (hazard ratio 0.695, 95% CI 0.559-0.864; P=.001), which remained significant after accounting for competing mortality (hazard ratio 0.578, 95% CI 0.484-0.691; P<.001). We selected the random forest model as the optimal model for the phenotypes, demonstrating high accuracy (area under the curve 0.914, 95% CI 0.911-0.918 for training and 0.867, 95% CI 0.858-0.876 for validation) and calibration ability (Brier score 0.111, 95% CI 0.109-0.113 for training and 0.132, 95% CI 0.127-0.137 for validation). CONCLUSIONS: By integrating unsupervised and supervised ML methods, we identified cardiovascular aging-related phenotypes that demonstrate robust predictive ability for incident stroke, which may have the potential to improve preventive and therapeutic strategies for high-risk populations.
2 May 2026
Read appraisal →Physics in medicine and biology
Few shots transfer learning for universal SPECT denoising across diverse acquisition protocols
Objective.Accelerated single photon emission computed tomography (SPECT) imaging, achieved by reducing either the number of projection angles or the acquisition time per angle, enhances clinical workflow efficiency but introduces elevated noise. This study aims to develop and validate a universal DL-based reconstruction framework that effectively generalizes across diverse, clinically-realistic SPECT acceleration protocols by overcoming the data scarcity challenge.Approach.SPECT bone scans from 103 patients were acquired under a standard scan (60 views, 12 s/view, 60v12s) followed by a fast scan using one of five acceleration protocols (60v6s, 60v3s, 30v12s, 30v6s, 16v12s). AU-Net-based reconstruction framework was implemented using three strategies: (1) single model, trained on individual acceleration protocol), (2) base model, trained on aggregated datasets from five acceleration protocols, and (3) transfer model, fine-tuned from the base model for protocol-specific optimization. Pixel-level accuracy and structural similarity were assessed using peak signal-to-noise ratio (PSNR), structural similarity index (SSIM) metrics, and maximum pixel value of lesions. Clinical evaluation of image quality, radionuclide detail, artifacts, and diagnostic confidence was conducted using a 5-point system.Main results.Quantitative evaluation showed the transfer model achieved better PSNR and SSIM across all protocols (highest 48.02 PSNR and 0.9918 SSIM in 30v6 s protocol). Qualitative analysis confirmed enhanced structural fidelity. Clinical evaluations rated the transfer model highest across metrics, with scores of 4.667 ± 0.508 (image quality, 30v12 s), 4.800 ± 0.250 (radionuclide detail, 60v6 s), 1.150 ± 0.173 (artifact reduction, 60v6 s), and 4.800 ± 0.250 (diagnostic confidence, 60v6 s), surpassing full-scan results in most cases.Significance. The proposed transfer learning (TL) framework effectively addressed data scarcity and improved reconstruction performance across diverse SPECT acceleration scenarios. The adoption of a TL strategy mitigates data scarcity by utilizing shared features and fine-tuning for specific protocols. The framework demonstrates potential for integration into fast SPECT workflows, facilitating reliable use across diverse imaging scenarios.
2 May 2026
Read appraisal →Neurology
Association of MRI-Visible Perivascular Spaces With Longitudinal Cognitive Decline Over a Decade
BACKGROUND AND OBJECTIVES: Cerebral small vessel disease (SVD) is the most common vascular contributor to dementia. SVD markers often coexist, contributing to difficulty assessing their independent contributions to cognitive domains. MRI-visible perivascular spaces (PVSs) are an emerging SVD marker visualized by MRI. We previously showed that basal ganglia PVSs cross-sectionally contribute to worse cognition, independent of other SVD markers. To further characterize the clinical relevance of PVS, we studied its role as a unique SVD marker of longitudinal cognitive decline. METHODS: Participants without stroke or dementia were included in the Vanderbilt Memory and Aging Project, a longitudinal observational cohort study based in Nashville, TN. Participants completed 3T MRI at study entry to measure SVD burden (PVS volume fraction, white matter hyperintensities volume, lacune counts, and cerebral microbleeds counts). PVS volumes were segmented using a deep learning algorithm. Participants underwent comprehensive serial neuropsychological testing over an 11-year follow-up period (mean follow-up = 4.9 ± 3.1 years). Each SVD marker was related to longitudinal neuropsychological performances using a linear mixed-effects model adjusting for age, sex, race/ethnicity, education, baseline cognitive status, apolipoprotein E-ε4 presence, Framingham Stroke Risk Profile, and intracranial volume. Head-to-head comparisons simultaneously tested multiple statistically significant SVD markers. RESULTS: Among participants (n = 750, age 68 ± 9 years, 52% female), higher basal ganglia PVS burden at baseline was associated with worse longitudinal performances in Boston Naming Test (β = -29.63; 95% CI -56.66 to -2.60), Animal Naming (β = -33.09; 95% CI -65.66 to -0.51), Wechsler Adult Intelligence Scale IV Coding (β = -86.36; 95% CI -150.9 to -21.82), executive function composite (β = -9.51; 95% CI -14.33 to -4.68), Hooper Visual Organization Test (β = -26.06; 95% CI -50.53 to -1.59), and episodic memory composite (β = -7.05; 95% CI -11.9 to -2.21). In head-to-head comparisons, basal ganglia PVS remained independent associations with executive function composite (β = -7.47; 95% CI -12.84 to -2.10) and Hooper Visual Organization Test (β = -22.11; 95% CI -42.38 to -1.85). DISCUSSION: Basal ganglia PVS burden independently contributes to worse longitudinal executive function and visuospatial skills independent of other SVD markers, highlighting PVS as an emerging marker of domain-specific cognitive decline in aging. Although causation cannot be established, findings further support PVS as a vascular contributor to deep brain structure damage underlying cognitive decline over time.
28 Apr 2026
Read appraisal →Biomedical physics & engineering express
BiFUSE-UNet: a novel dual-encoder enhanced TransUNet for polyp segmentation
Polyps are abnormal protrusions that develop on the mucosal surface of the human body, arising from hyperproliferation or aberrant differentiation of mucosal tissue. While some polyps carry a risk of malignant transformation, timely intervention can halt this progression. Accurately segmenting polyp contours in medical images poses significant challenges, primarily due to the high heterogeneity in polyp appearance and morphology, blurry boundaries with background interference, and class imbalance coupled with data scarcity. To address these challenges, this paper presents BiFuse-UNet, a novel dual-encoder TransUNet architecture built upon a bidirectional fusion mechanism. Building upon the original TransUNet architecture, BiFuse-UNet incorporates several key modules. Firstly, a multi-axis attention module (MaxViT block) is integrated to form a dual-encoder structure, which extracts multi-scale local textural features across different dimensions and captures long-range dependencies through self-attention. Secondly, a Merge fusion module is designed, incorporating a Coordinate Attention module and the SaE block from SENetV2, to enhance the model's perception of local and detailed features. Finally, a Biformer module is introduced to strike an efficient balance between global semantics and local details, thereby improving the segmentation accuracy for subtle polyps and early-stage polyps with atypical morphology. Extensive evaluations on the Kvasir-SEG and CVC-ClinicDB dataset demonstrate that BiFuse-UNet achieves state-of-the-art segmentation performance, outperforming seven existing methods on key metrics such as the Dice coefficient, Recall, Precision, and IoU. The proposed model effectively mitigates the challenges of morphological heterogeneity, blurry boundaries, and background interference, leading to highly accurate pixel-level segmentation.
27 Apr 2026
Read appraisal →Physics in medicine and biology
Confounding factors in the generalisation of synthetic CT to diagnostic spine MRI
Objective. Generalisation of synthetic CT (sCT) generation to diagnostic MRI data in the spine faces many challenges, particularly when aiming at accurate visualisation of pathologies for orthopaedic treatment management. In this study we assessed the effect of potential confounding factors on the performance of sCT generation from diagnostic MR images.Approach. Paired spinal diagnostic MR and CT scans from two centres (51 patients) were collected retrospectively, spanning multiple spinal pathologies and regions. Each patient's dataset contained 3D T1-/T2-weighted and 2D T1- and T2-weighted scans. 3D U-Nets were trained per centre on the 3D MR data and CT using a mean absolute error (MAE) loss. The performance was assessed on sCTs derived from original 2D and 3D data and from simulated 2D data with varying orientation (axial/coronal/sagittal) and slice spacing (1.1-6.6 mm). The sCTs were compared to the CTs using MAE, peak signal-to-noise ratio, and dice similarity coefficient.Main results. We identified MR resolution, orientation and contrast as confounding factors for sCT generation from diagnostic MR. Additionally, CT noise and CT-to-MR registration were identified to influence the performance evaluation. Performance degraded with larger slice spacings and significant differences between orientations occurred more often at larger slice spacings. The effect of slice spacing and orientation was stronger inside the vertebrae than outside. While the networks performed better overall on contrasts more similar to the training contrasts, they demonstrated promising performance when trained on 3D T2w images and tested on 2D T1w images. Finally, higher CT noise levels resulted in worse sCT performance metrics, reflected in a significant correlation between noise and MAE.Significance. This work demonstrates that generalisation of sCT generation to diagnostic spine MRI data is hampered primarily by MRI acquisition related aspects including MR resolution, orientation and contrast, which provides guidance for future work on generalisable sCT from diagnostic MR.
27 Apr 2026
Read appraisal →Physics in medicine and biology
MRDT-GAN: generative adversarial network with multi-scale residual dense transformer generator for low-dose CT denoising
Objective.Low-dose computed tomography (LDCT) reduces radiation exposure but introduces noise and artifacts that degrade diagnostic quality. Existing deep learning-based denoising methods still face challenges such as over-smoothing, loss of fine structures, and uneven contrast. This study aims to develop an LDCT denoising framework that enhances noise suppression while preserving anatomical details and structural fidelity.Approach.We propose a multi-scale residual dense transformer generative adversarial network (MRDT-GAN). In the generator, we adopt the MRDT Block as the core unit, which introduces multi-scale strategy into residual dense network to reduce over-smoothing and preserve fine details, and also patching transformer block to capture long-range dependencies, mitigating distortions caused by localized receptive fields in convolutional neural network-based approaches. A hybrid attention module is also introduced in the generator to process spatial, frequency, and contrast information, enabling the network to focus on critical regions for noise suppression, improve contrast uniformity, and maintain texture consistency. In the discriminator, we adversarially explore differences on global, pixel, and also sub-scale between denoised LDCT and normal dose CT to better capture structural variations, reduce local noise and distortions, and ensure more realistic texture reconstruction while minimizing artifacts.Main results.We validate MRDT-GAN on both the NIH-AAPM-Mayo Clinic LDCT dataset and a real-world dataset. Experimental results indicate that MRDT-GAN achieves superior denoising performance compared with existing methods, effectively preserves details, enhances visual quality, and achieves a better balance between noise suppression and structural integrity.Significance.MRDT-GAN provides an effective and generalizable LDCT denoising solution that balances noise reduction with fine-detail preservation. By integrating MRDT modeling, hybrid attention mechanisms, and multi-difference adversarial learning, the framework offers improved clinical applicability and supports high-quality image reconstruction for downstream diagnostic tasks.
27 Apr 2026
Read appraisal →Annals of medicine
Construction of an XGBoost-SHAP-based malignant transformation risk prediction model for gallbladder polyps
BACKGROUND: To develop and validate a risk prediction model for malignant transformation in patients with gallbladder polyps (GBPs) using an interpretable machine learning framework and evaluate its predictive accuracy. METHODS: A retrospective cohort of 1,027 surgical patients was enrolled from Yantai Yuhuangding Hospital (training set: n = 933) and Shanghai Eastern Hepatobiliary Surgery Hospital (validation set: n = 94). Feature selection for the training set was performed using the least absolute shrinkage and selection operator (LASSO) regression method. A predictive model was constructed with the XGBoost machine learning algorithm and evaluated using Shapley Additive exPlanation (SHAP). RESULTS: LASSO regression identified five significant risk factors for malignant transformation in GBPs: presence of concomitant cholecystitis, polyp count, polyp base width, age, and maximum polyp diameter. The area under the receiver operating characteristic curve (AUC) was 0.862 (95% confidence interval [CI]: 0.8342-0.8893) in the training set and 0.777 (95% CI: 0.6804-0.8737) in the validation set. SHAP analysis illustrated the contribution of each factor. CONCLUSION: This study developed and validated a risk prediction model for malignant transformation in patients with GBPs. The model demonstrated favorable discrimination, calibration, accuracy, and clinical applicability. Integration with SHAP technology may assist clinicians in optimizing treatment and management strategies.
27 Apr 2026
Read appraisal →Reviews in the neurosciences
Microstructural alterations of the amygdala in neurodegenerative and neuroinflammatory disorders: insights from diffusion tensor imaging
Diffusion tensor imaging (DTI) is a valuable method for evaluating microstructural changes in the amygdala associated with neurodegenerative and neuroinflammatory disorders. This systematic review examines amygdala microstructural alterations in Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), dementia with Lewy bodies (DLB), and multiple sclerosis (MS) using DTI metrics. Following PRISMA 2020 guidelines, we searched PubMed, Scopus, and Web of Science databases through August 2025, identifying 4,442 records. After screening and eligibility assessment, 13 studies were included, comprising 1,412 patients and 1,146 healthy controls. Due to sample heterogeneity and lack of standardized effect size measures, meta-analysis was not performed. Across disorders, elevated mean diffusivity (MD) emerged as the most consistent finding, present in 100 % of AD patients and observed in all examined conditions. Reduced fractional anisotropy (FA) was the second most frequent alteration, with 36.6 % of AD patients showing decreased FA. In MS, increased radial diffusivity (RD) was prominent, while longitudinal DLB studies revealed progressive free water (FW) increases. These DTI-based microstructural changes often preceded volumetric atrophy and correlated with clinical severity. Our findings demonstrate that DTI metrics, particularly MD and FA, serve as sensitive markers of amygdala pathology across neurodegenerative diseases and may facilitate early diagnosis, disease monitoring, and differential diagnosis of these conditions.
26 Apr 2026
Read appraisal →Biomedizinische Technik. Biomedical engineering
Advanced deep learning framework for breast cancer detection using digital breast tomosynthesis images
OBJECTIVES: Early and accurate detection of breast cancer is critical for improving patient survival outcomes. This study proposes a robust deep learning-based framework for breast cancer detection using digital breast tomosynthesis (DBT) images, leveraging both single-slice and multi-slice inputs. METHODS: The proposed work includes image normalization, resizing and Laplacian Pyramid Enhancement (LPE). Various features were extracted and fused in different combinations. To retain the most discriminative features Exhaustive Feature Selection (EFS) is used. A hybrid model integrated using ResNet V2, MobileNet V3 and Inception V3+ for classification. Finally ensemble learning with XGBoost was applied and Hyperparameters were optimized using a grid search strategy. RESULTS: The hybrid model with multi-slice DBT inputs achieved better improvements with accuracy, sensitivity, specificity and area under the curve (AUC). While applying LPE, feature fusion and EFS substantially enhanced the hybrid model's diagnostic performance. CONCLUSIONS: The findings demonstrate the strong potential of advanced methods in enhancing performance. Future research work will focus on integrating this pipeline with clinical decision support systems, multi-center datasets and extending to other breast imaging modalities.
26 Apr 2026
Read appraisal →Biomedical physics & engineering express
Impact of multimodal information on the estimation of fetal growth indicators using machine learning regression models
Accurate estimation of fetal growth indicators such as birth weight, birth length, and gestational age at birth is essential for monitoring pregnancy outcomes and guiding clinical decisions. Traditional predictive models typically rely on ultrasound-based fetometric data to estimate fetal weight or length. While valuable, these models provide estimates only at the time of measurement, rather than predicting values at birth, and may overlook important clinical and sociodemographic factors that also influence fetal growth. This study aimed to evaluate whether incorporating echographic, clinical, and sociodemographic features could improve the accuracy of predicting fetal growth indicators at birth and to quantify the contribution of each variable. Data from 154 cases were collected and processed for model development (61.5% for training and 38.5% for testing), divided into three feature sets: fetometric, clinical-sociodemographic, and combined clinical, echographic, and sociodemographic data. Six regression models were developed to predict three fetal growth indicators: birth weight, birth length, and gestational age at birth. Model performances were assessed usingR2, mean absolute error (MAE), and mean absolute percentage error (MAPE). The multimodal models significantly outperformed those relying only on fetometric or clinical-sociodemographic data, with the random forest achieving the best performance for birth weightR2: 0.8991; MAE: 255.08 g; MAPE: 8.46%, birth lengthR2: 0.8679; MAE: 7.21 cm; MAPE: 2.73%, and gestational age at birthR2: 0.8886; MAE: 1.34 d; MAPE: 2.76%. Feature relevance analysis revealed that variables such as maternal height, maternal weight, placenta location, and alcohol consumption played substantial roles in prediction accuracy, alongside classic fetometric measurements such as head circumference. These findings highlight the multifactorial nature of fetal growth and demonstrate that integrating clinical and sociodemographic information enhances the performance of fetal growth prediction models, ultimately supporting improved perinatal care.
25 Apr 2026
Read appraisal →La Medicina del lavoro
Artificial Intelligence in Occupational Health Surveillance: Evaluating AI-Assisted ILO Classification of Radiographs of Pneumoconioses
BACKGROUND: Pneumoconioses remain an important occupational health issue, particularly in low- and middle-income countries. The International Labour Organization (ILO) Classification standardizes chest radiograph interpretation but requires trained readers and is affected by inter-reader variability. This study evaluated whether generative multimodal artificial intelligence (AI) models can approximate ILO-based diagnostic reasoning. METHODS: Eighty-two chest radiographs from the official NIOSH B Reader syllabus were analysed using four AI systems (GPT-4o, GPT-5, MedGemma-4B, MedGemma-27B). Each image was evaluated with a standardized prompt based on the 2022 revised ILO guidelines using deterministic settings. Model outputs were mapped to ILO codes and compared with the official answer keys of the ILO Standard Radiograph Set used for B Reader training and examination. Performance metrics included balanced accuracy, sensitivity, specificity, precision, and Matthews correlation coefficient (MCC). Bootstrap 95% confidence intervals, McNemar's test, and Cohen's κ assessed performance variability and agreement. RESULTS: All four AI models showed moderate diagnostic performance, with balanced accuracy ranging from 60.8% to 70.3%. Sensitivity remained limited (35.5%-54.9%), while specificity was consistently high (84.6%-86.2%). MedGemma-27B performed best for small opacities, GPT-5 for pleural abnormalities and for technical quality. Large opacities and rare findings were systematically under-detected. Statistical comparisons showed significant differences between models, although agreement patterns were broadly similar. CONCLUSION: All AI models partially followed structured ILO radiographic criteria but did not achieve expert-level performance, confirming that they cannot replace certified B Readers. Larger, real-world datasets are needed to assess their potential clinical utility as supportive tools in occupational health surveillance programs.
25 Apr 2026
Read appraisal →Biomedical physics & engineering express
A graph deep learning method for diagnosis of Parkinson's disease using brain functional connectivity features
Early and precise identification of Parkinson's disease (PD) is crucial for clinical intervention. Resting-state functional magnetic resonance imaging (rs-fMRI) provides a valuable approach for revealing PD-related differences in brain functional connectivity (FC). However, existing methods often focus solely on characterizing the spatial topology of FC while neglecting its time-varying dynamic fluctuations. Furthermore, they frequently exhibit limited generalization capability when dealing with small sample sizes, and their decision-making mechanisms lack interpretability. To address these limitations, this study proposes an interpretable graph convolutional network framework. This framework integrates both static and dynamic FC information to capture both the stable topological structure and the dynamic temporal characteristics of brain networks. Simultaneously, it models population relationships by constructing an inter-subject similarity graph to enhance the model's representational capacity. Additionally, this study incorporates interpretability analysis techniques to deeply dissect the model's decision-making mechanism and identify key brain regions critical for classification. Results demonstrate that the proposed model achieves superior performance in PD classification tasks and exhibits good generalization ability. More importantly, by interpreting the model's decisions, key brain regions associated with PD discrimination were successfully identified. This study provides an effective computational framework for PD identification and offers new insights into understanding its pathological mechanisms.
25 Apr 2026
Read appraisal →Biomedical physics & engineering express
Evaluating the impact of reconstruction algorithms on digital breast tomosynthesis system optimization for microcalcification detection
Early detection of microcalcifications in digital breast tomosynthesis (DBT) depends on optimizing both image reconstruction algorithms and system design parameters. Thisin-silicostudy evaluates the combined effects of x-ray tube motion, angular dose distribution, and detector characteristics on reconstruction algorithms for microcalcification detectability. A synthetic breast phantom containing 127 microcalcification clusters was generated and imaged under four acquisition conditions varying in tube motion (continuous vs. step-and-shot), angular dose distribution (uniform vs. non-uniform), and detector performance (noise level and pixel size). Projection data were reconstructed using filtered back-projection (FBP) and the simultaneous algebraic reconstruction technique (SART). Detection performance was assessed using a 2D Filtered Channel Observer in a multi-reader, multi-case study. Evaluation metrics included signal-to-noise ratio and area under the receiver operating characteristic (ROC) curve. Results show that SART performs best under step-and-shot acquisition by minimizing motion blur and enhancing fine detail, while FBP benefits more from non-uniform angular dose distributions. Detector improvements-lower noise and finer resolution-enhance detectability across all configurations, with FBP showing the largest gains. These findings highlight the interdependence of reconstruction algorithm, tube motion, dose strategy, and detector design in DBT system optimization. This study provides a detailed investigation into classical reconstruction strategies and establishes a methodological baseline for future DBT optimization research, offering actionable guidance for acquisition protocol design and hardware development.
24 Apr 2026
Read appraisal →PloS one
LHW-Net: An ensemble-based machine learning framework for brain tumor classification
The classification of brain tumors is an unsolved problem associated with heterogeneity of tumors and fluctuations in imaging conditions. In this work, the investigation introduces a powerful novel framework, named LHW-Net that combines handcrafted features called local binary patterns (LBP), histogram of oriented gradients (HOG), and wavelet transform (WT). Within the LHW-Net framework, the extracted features are utilized in different machine learning classifiers, such as K-Nearest Neighbors (KNN), Random Forest (RF) and Support Vector Classifier (SVC). The results of the individual classifiers are further combined using probabilistic score fusion approach to improve classification performance. The effectiveness and robustness of the proposed work are validated by the achieved experimental results on commonly accepted benchmark datasets.
24 Apr 2026
Read appraisal →The Journal of clinical endocrinology and metabolism
An Interpretable Machine-Learning Model for Predicting Occult Central Lymph Node Metastasis in Papillary Thyroid Cancer
CONTEXT: Accurate preoperative prediction of occult lymph node metastasis (OLNM) in clinically lymph node negative (cN0) papillary thyroid carcinoma (PTC) is critical for optimizing therapeutic strategy, particularly for thermal ablation and active surveillance. OBJECTIVE: The aim of this study was to develop an interpretable machine-learning (ML) model to predict the risk of OLNM in cN0 PTC patients. METHODS: This retrospective study analyzed data of 961 cN0 PTC patients (August 2018-August 2023). Multivariable logistic regression identified independent risk factors for OLNM in cN0 PTC. The cohort was randomly divided into the training and test sets, and a subset of patients with tumors sized 1 cm or less was further extracted from the test set for internal validation. Eight ML models incorporating clinical, ultrasonographic, and molecular features were developed and evaluated. Shapley Additive exPlanations (SHAP) enhanced interpretability. RESULTS: RET fusion positivity and BRAF mutation positivity were identified as independent molecular risk factors for OLNM in cN0 PTC, alongside 6 clinical and ultrasonographic variables. Nine predictors were incorporated into the predictive model. The random forest (RF) model achieved optimal performance with an area under the curve (AUC) of 0.906 in the training set and 0.733 in the test set, along with the lowest Brier scores of 0.135 and 0.212, respectively. Analysis of tumors sized 1 cm or less internally validated the model's robustness with an AUC of 0.719. SHAP analysis identified size, age, and clustered punctate echogenic foci as the top predictors. CONCLUSION: This is the first study to identify RET fusion positivity as an independent OLNM risk factor in cN0 PTC. The developed RF model demonstrates moderate predictive performance for OLNM risk and provides a framework for integrating clinical, sonographic, and molecular data, and is deployed as a web calculator (https://predictingoccultlymphnodemetastasis.shinyapps.io/web3/).
23 Apr 2026
Read appraisal →PloS one
Controlled comparative study of YOLOv8-Pose, YOLOv11-Pose, and Detectron2 for vertebrae detection and keypoint estimation
Accurate vertebrae detection with precise keypoint localization is essential for medical image analysis and clinical applications, such as spinal alignment assessment. This study presents a controlled, task-driven comparison of four pose-based deep learning models for vertebrae detection and keypoint estimation: YOLOv8n-Pose, YOLOv11n-Pose, and Detectron2 (Keypoint R-CNN) with ResNet-50 and ResNet-101 backbones. A single-class vertebrae dataset with bounding box annotations and four anatomically meaningful keypoints per vertebra was evaluated under consistent training settings. Performance was assessed in terms of keypoint localization accuracy, detection precision, inference speed, and vertebrae-specific prediction behavior, including detection completeness and duplicate detections, which are critical for clinical usability. Results show that YOLOv11n-Pose provides the best balance between keypoint accuracy and inference efficiency, while YOLOv8n-Pose achieves the fastest inference with competitive performance. Detectron2-based models exhibit lower keypoint accuracy, slower inference, and frequent duplicate predictions that obscure correct anatomical landmarks. Additional experiments indicate that larger YOLO variants (particularly YOLOv8l-Pose) improve accuracy, albeit at increased computational cost. Overall, the results demonstrate that model capacity or architectural recency alone does not guarantee superior performance in vertebrae keypoint detection, highlighting the importance of anatomy-aware model selection for spine imaging applications.
23 Apr 2026
Read appraisal →Medical science monitor : international medical journal of experimental and clinical research
Use of Deep Learning Models in the Diagnosis of Proptosis Through Orbital Magnetic Resonance Imaging
BACKGROUND Proptosis is a common manifestation of orbital disease; however, current diagnostic tools, such as the Hertel exophthalmometer and manual radiological measurements, have limited reproducibility and are observer-dependent. More objective, automated approaches are needed. In this single-center retrospective study, orbital magnetic resonance imaging (MRI) examinations from 521 participants (261 with proptosis, 260 controls) were analyzed. Proptosis was defined on MRI using interzygomatic line-based distance criteria. Three-dimensional convolutional neural network models based on DenseNet121, DenseNet169, DenseNet264, and ResNet50 architectures were trained on volumetric orbital MRI data. MATERIAL AND METHODS Data were divided into training, validation, and test sets, and 5-fold cross-validation with early stopping was used to optimize and validate model performance. Diagnostic performance was assessed using accuracy, area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. RESULTS DenseNet121 achieved the best overall performance, with mean accuracy of 95.0%, AUC of 0.986, sensitivity of 92.7%, and specificity of 96.9% across 5-fold cross-validation. CONCLUSIONS To the best of our knowledge, prior artificial intelligence studies in orbital imaging have primarily focused on CT-based measurements, radiomics approaches, or thyroid-associated orbitopathy assessment rather than end-to-end 3-dimensional deep learning analysis of orbital MRI volumes. In this context, the present study explores a volumetric MRI-based deep learning framework for automated proptosis detection, emphasizing patient-level classification and explainable model interpretation.
19 Apr 2026
Read appraisal →BMJ open
Prevalence of incidental findings requiring interventions in CT for trauma patients in the emergency department: a systematic review and meta-analysis
OBJECTIVE: To systematically review and meta-analyse the prevalence of incidental findings (IFs) requiring interventions identified in CT performed for patients with traumatic injuries in the emergency department, including pathologically confirmed cancers and emergent non-traumatic vascular pathologies. DESIGN: Systematic review and meta-analysis. DATA SOURCES: PubMed, EMBASE and Cochrane CENTRAL databases from database inception to 22 November 2024. ELIGIBILITY CRITERIA FOR SELECTING STUDIES: Prospective or retrospective studies involving traumatic injury patients presented at the emergency department reporting IFs of clinical significance detected through CT with any interventions proposed were included. Studies that exclusively included paediatric populations were excluded. The systematic review methods included double-screening, dual assessment of eligibility and study validity, dual data extraction, Bayesian multivariate random-effects meta-analysis of prevalence and employing the Grading of Recommendations, Assessment, Development and Evaluations' rating for the certainty of evidence. The primary outcomes were IFs requiring any interventions, and subset IFs requiring urgent interventions. Secondary outcomes were histologically confirmed cancers and emergent non-traumatic vascular pathologies. RESULTS: 22 studies (1 with a prospective and 21 with a retrospective design) mainly from high-income countries using limited-quality data based on clinical practice involving 18 538 patients were included. 9 studies evaluated the whole body, while 13 evaluated selective body regions. The grading criteria for IFs were non-uniform, and the image interpreters involved had diverse experience and expertise. The summary prevalence estimates for IFs requiring any interventions, urgent interventions, pathologically confirmed cancers and emergent non-traumatic vascular pathologies detected in the whole body were 29.8% (95% credible interval (CrI) 20.4% to 42.9%; very low certainty), 7.6% (95% CrI 4.5% to 14.8%; low certainty), 0.6% (95% CrI 0.3% to 1.6%; moderate certainty) and 0.3% (95% CrI 0.1% to 0.9%; moderate certainty), respectively. These findings were largely identified in the chest or abdomen and pelvis, with the overall detection frequency reduced with the scanned body regions narrowed (very low to moderate certainty). Sparse data on the head, neck and spine resulted in limited results. CONCLUSIONS: IFs identified in trauma whole-body CT requiring intervention are prevalent and can lead to substantial medical costs. The widely reported prevalence range suggests variations in radiologist recommendations and reporting in clinical practice and calls for standardisations. IFs requiring urgent intervention are not rare, which leads to a diagnosis of significant diseases including cancers and urgent vascular pathologies. Future studies should report long-term, patient-relevant results based on standardised classification and reporting systems. PROSPERO REGISTRATION NUMBER: CRD42020187852.
18 Apr 2026
Read appraisal →Acta orthopaedica
Migration patterns of uncemented femoral stems in hip replacement: a systematic review and meta-analysis of clinical radiostereometric analysis cohort studies
BACKGROUND AND PURPOSE: We conducted a systematic review and meta-analysis of radiostereometric analysis studies of primary uncemented femoral stems to investigate their subsidence and retroversion patterns and the migration patterns according to implant, patient, and study characteristics. METHODS: A systematic search of PubMed, Web of Science, Cochrane, and Embase databases to identify all radiostereometric analysis studies on femoral stem migration following primary uncemented total hip replacement was performed. Clinical studies with 2 or more postoperative radiostereometric measurements within 2 years were included. Subsidence and retroversion at 6 weeks, 3 and 6 months, 1, 2, 5, and 10 years were included for analysis. Extracted implant characteristics included implant design, coating, and surgical approach. Data was analyzed using a random effects model. RESULTS: 73 studies on 120 cohorts and 2,970 uncemented stems were included. 119 cohorts reported on subsidence and 91 on retroversion. The pooled subsidence at 3 months was 0.29 mm (95% confidence interval [CI] 0.19-0.39) and 0.32 mm (CI 0.21-0.43) at 2 years. The pooled retroversion at 3 months was 0.51° (CI 0.33-0.70) and 0.70° (CI 0.48-0.93) at 2 years. Hydroxyapatite-coated stems showed the least migration (subsidence 0.26 mm; CI 0.13-0.40; retroversion 0.51°; CI 0.22-0.80) among different coating types. The anterior approach showed more migration (subsidence 1.04 mm, CI 0.53-1.55; retroversion 1.52°, CI 1.08-1.95) than other surgical approaches. CONCLUSION: Our study shows that most subsidence and retroversion of uncemented femoral stems occurs during the first 3 months. Stabilization of subsidence occurred after 3 months, and retroversion stabilized after 2 years. Migration patterns differ based on stem type, coating, surgical approach, the time period when inclusion started, and timing of baseline measurement.
18 Apr 2026
Read appraisal →PloS one
Statistically valid explainable black-box machine learning: applications in sex classification across species using brain imaging
Sex classification using neuroimaging data has the potential to revolutionize personalized diagnostics by revealing subtle structural brain differences that underlie sex-specific disease risks. Despite the promise of machine learning, traditional methods often fall short in providing both high classification accuracy and interpretable, statistically validated feature importance scores for high-dimensional imaging data. This gap is particularly evident when conventional techniques such as random forests, LIME, and SHAP are applied, as they struggle with complex feature interactions and managing noise in large datasets. We address this challenge by developing an integrated framework that combines Oblique Random Forests (ORFs) with a novel, permutation-based feature importance testing algorithm. ORFs extend traditional random forests by employing oblique decision boundaries through linear combinations of features, thereby capturing intricate interactions inherent in neuroimaging data. Our feature importance testing method, NEOFIT, rigorously quantifies the significance of each feature by generating null distributions and corrected p-values. We first validate our approach using simulated datasets, establishing its robustness and scalability under controlled conditions. We then apply our method to classify sex from both voxel-wise structural MRI and cortical thickness data in humans and macaques, facilitating direct cross-species comparisons. ORFs achieves AUC > 0.80 on human data, and >0.70 on macaque data, while NEOFIT identifies statistically significant features aligned with sex-dimorphic neuroanatomy. Our results demonstrate that the proposed framework not only enhances classification performance but also provides clear, interpretable insights into the neuroanatomical features that distinguish sexes. These methodological advancements pave the way for improved diagnostic tools and contribute to a deeper understanding of the evolutionary basis of sex differences in brain structure.
18 Apr 2026
Read appraisal →BMJ open
Indocyanine green fluorescence for intraoperative detection of liver tumours in minimally invasive surgery: protocol for the LIVERGREEN phase IV multicentre clinical trial
INTRODUCTION: Liver tumours are a leading cause of global morbidity and mortality. Current diagnostic tools, including computed tomography (CT), magnetic resonance imaging (MRI) and intraoperative ultrasound (IOUS), have limitations in detecting liver neoplasms. Indocyanine green (ICG) has emerged as a promising tool for improving liver tumour detection. This study aims to assess the impact of preoperative ICG on intraoperative tumour detection in minimally invasive surgery and develop a machine-learning algorithm to enhance tumour detection using ICG fluorescence. METHODS AND ANALYSIS: This prospective, multicentre, phase IV clinical trial adheres to Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines. Patients with liver tumours eligible for minimally invasive surgery and a preoperative imaging test will be included. ICG will be administered intravenously 24 hours before surgery. Intraoperative procedures will include IOUS, ICG mapping and photographic documentation. Patients will be followed for 90 days to assess tumour progression, morbidity and mortality. The photographic analysis will enable the development of an artificial intelligence algorithm using machine learning and neural networks to identify lesions based on ICG fluorescence. The estimated sample size is 173 patients and the trial is predicted to accrue in 3 years. ETHICS AND DISSEMINATION: The trial will be conducted in accordance with the Declaration of Helsinki and the Spanish Agency of Medicines and Medical Devices (AEMPS) guidelines. Approved by the local institutional Ethics Committee and the AEMPS, the results will be shared with the scientific community through publications and conferences. EUDRACT NUMBER: 2023-5 08 316-27-00. TRIAL REGISTRATION NUMBER: NCT06398028. PROTOCOL VERSION: V.12, 18 March 2025.
18 Apr 2026
Read appraisal →PloS one
LungNet: Leveraging state-space models with SE-enhanced skip connections for precise CT-based lung lesion segmentation
Lung cancer, which accounted for 2.48 million new cases and 1.82 million deaths worldwide in 2022, continues to be the most lethal cancer across the globe, underscoring the urgent demand for more advanced diagnostic tools. Although computed tomography (CT) imaging has long been central to lung cancer detection, the heterogeneous and complex characteristics of lung lesions make accurate segmentation particularly challenging. Current deep learning methods face a critical bottleneck: convolutional neural networks (CNNs) often struggle to capture long-range dependencies due to limited receptive fields, while Transformer-based architectures incur prohibitive computational costs when processing high-resolution CT volumes. Furthermore, standard skip connections in traditional UNet models frequently introduce redundant noise, leading to the dilution of subtle lesion features. To address these specific technical gaps, we introduce a novel deep learning framework that integrates Mamba state-space models with an improved UNet architecture. In this design, to mitigate feature redundancy, Squeeze-and-Excitation networks are embedded into skip connections, while auxiliary losses are introduced to address the degradation of shallow features and capture fine-grained lesion features across varied morphologies. Such a framework not only accommodates the intricate differences in lesion size, shape, and spatial distribution but also achieves a balance between global context modeling and linear computational efficiency. By uniting the local feature extraction strengths of convolutional layers with the long-range dependency modeling power of state-space models, our approach achieves more precise delineation of lung lesion boundaries. Extensive experiments conducted on multiple datasets provide compelling evidence of the method's effectiveness: it attains state-of-the-art segmentation accuracy and demonstrates significant promise for enhancing early detection, ongoing disease monitoring, and treatment planning in lung cancer patients. This advancement delivers a robust solution for the inherently complex task of lung lesion segmentation. Moreover, because lung cancer treatment is costly and insurance coverage plays a decisive role in distributing expenses, the study's outcomes also carry considerable implications for the insurance sector.
17 Apr 2026
Read appraisal →Spine
An Unsupervised Learning Approach for Multimodal Low Back Pain Stratification
STUDY DESIGN: Cross-sectional study. OBJECTIVE: This study proposes a novel stratification framework for individuals with low back pain (LBP). The method integrates Northern Finland Birth Cohort data comprising imaging biomarkers from deep learning (DL)-based analysis of lumbar spine MRI with the data on smoking status, demographics (sex and BMI), self-reported data from Örebro Musculoskeletal Pain Screening Questionnaire (ÖMPSQ) short and the STarT Back Tool (SBT). Furthermore, the utility of this stratified approach was validated by demonstrating a superior net benefit compared with "treat-all" strategy. BACKGROUND: Current risk stratification for individuals with LBP relies on ÖMPSQ short and SBT among others. While these tools are invaluable for capturing psychosocial characteristics predictive of future disability and functional outcomes, LBP's multifactorial nature necessitates a more comprehensive framework for effective risk stratification. MATERIALS AND METHODS: A method for multimodal unsupervised patient stratification has been developed that allows for the integration of imaging biomarkers of disc degeneration (DD) and facet tropism (FT), extracted using DL models, with nonimaging data. The framework utilized robust K-Means clustering to stratify individuals. Clusters were characterized using LBP frequency and bothersomeness, and their robustness was validated with a multi-class logistic regression model. Net benefit was assessed through decision curve analysis. RESULTS: Three distinct subgroups were characterized by LBP frequency and bothersomeness. One subgroup was dominated by psychosocial characteristics (psychosocial risk P < 0.05), the second by physical degenerative changes (DD P < 0.05), and the third by a mix of both. Predictive models for cluster assignment were robust, achieving high mean accuracies (SBT-based: 0.89; ÖMPSQ-short-based: 0.87). The net benefit is superior throughout a range of threshold probabilities compared with a "treat-all" strategy. CONCLUSION: A novel framework was developed that integrates multimodal data to identify distinct subgroups differentiated by their physical and psychosocial characteristics in a population-based cohort, demonstrating potential for advancing personalized care.
17 Apr 2026
Read appraisal →Theory in biosciences = Theorie in den Biowissenschaften
Analytical and AINN-based investigation of surface wave propagation in dry long bones with initial stress, magnetic field, and rotation effects
This study presents a comprehensive analytical and artificial intelligence neural network (AINN)-based investigation of wave propagation in dry long bones, modeled as an orthotropic hollow cylindrical structure. The proposed model incorporates the combined effects of initial stress, magnetic field, and rotational motion to capture the complex behavior of bone-like media under coupled physical influences. The governing equations are formulated within a continuum mechanics framework and solved analytically using displacement potential methods, yielding solutions in terms of Bessel functions that satisfy the cylindrical geometry and boundary conditions. Two distinct cases, corresponding to the absence and presence of rotation, are examined to assess the influence of rotational effects on wave dynamics. A detailed parametric analysis is carried out to evaluate the variation of phase velocity and frequency with respect to wave number, initial stress, magnetic parameter, density, and geometric ratios. The analytical results indicate that initial stress enhances wave propagation, while magnetic effects introduce damping and rotation significantly modifies dispersion characteristics. To improve computational efficiency and predictive capability, an AINN model is developed and trained using analytically generated data. The AINN predictions show excellent agreement with the analytical results, as confirmed through parity plots, error analysis, residual distribution, and loss convergence behavior. The novelty of the study lies in the unified analytical-AINN framework that integrates mechanical, electromagnetic, and rotational effects within a single model. The findings provide important insights for wave-based characterization of bone structures and have potential applications in non-destructive evaluation, ultrasonic diagnostics, and advanced biomedical sensing systems.
17 Apr 2026
Read appraisal →Journal of the American College of Surgeons
Predictive Model for Mild Autonomous Cortisol Secretion in Patients Evaluated in Artificial Intelligence-Driven Adrenal Incidentaloma Clinic
BACKGROUND: Mild autonomous cortisol secretion (MACS) is present in approximately 20% to 50% of adrenal incidentalomas. These patients do not exhibit the clinical manifestations of overt Cushing's syndrome, and differentiation from nonfunctional adrenal incidentalomas (NFAI) is typically made after a low-dose dexamethasone suppression test. The objective of this study was to develop predictive models to distinguish MACS from NFAI using clinical and radiological parameters. STUDY DESIGN: This retrospective study included patients evaluated in an adrenal incidentaloma clinic between February 2022 and August 2024 who were diagnosed with either NFAI or MACS. Demographic characteristics, medical and medication history, and radiological features were collected. Patients were randomly divided into training and test cohorts in a 3:1 ratio. Predictive models for MACS were developed using Least Absolute Shrinkage and Selection Operator (LASSO) regression and random forest (RF) algorithms. RESULTS: A total of 397 patients were included, with 297 allocated to the training set and the remaining 100 to the test set. The mean age was 62.3 years, and 55% of participants (n = 220) were women. MACS was present in 34% of the study population (n = 136). The most influential predictive factors of MACS were BMI, Posterior Adiposity Index, and the number of antihypertensive medications. The LASSO and RF models achieved discrimination with area under the curve values of 0.686 and 0.736, respectively. At Youden Index thresholds balancing sensitivity and selectivity, the LASSO model had 58.8% sensitivity, 75.8% specificity, and 70% accuracy, whereas the RF model had 64.7% sensitivity, 75.8% specificity, and 72% accuracy. CONCLUSIONS: Predictive models incorporating clinical and radiological characteristics offer a promising approach for distinguishing MACS from NFAI in patients with adrenal incidentalomas.
17 Apr 2026
Read appraisal →Computers in biology and medicine
Empirical evaluation of variability and multi-institutional generalizability of deep learning survival models: application to renal cancer CT scans
Deep learning survival (DLS) modeling has shown significant promise across multiple disease conditions for predicting patient outcomes via biomedical data, but remain relatively under-explored in conjunction with medical imaging. This poses the question of what methodological choices promote model generalization, especially when predicting continuous survival outcomes in external validation cohorts. In this work, we systematically examined the impact of data partitioning, data order, model initialization, and augmentation strategies on the generalization of CT-based DLS survival models in renal cancers. A multi-institutional CT cohort was assembled, leveraging public repositories such as TCGA-KIRC, TCGA-KICH, TCGA-KIRP, and KITS19; totaling 525 patients from across 9 institutions. A 3D ResNet-18 model was trained to predict overall patient survival with a Cox loss function and evaluated across three controlled experiments: (i) random versus covariate-balanced data partitioning, (ii) randomizing vs fixing data ordering and initial model weights, and (iii) varying intensity, spatial, and noise based augmentations; evaluated in terms of concordance index (c-index) and hazard ratio (HR). Intelligent covariate-balanced data partitioning demonstrated optimal performance on an external validation set (c-index 0.74, HR 5.08) compared to random partitioning (c-index 0.56, HR 0.29). Different model initializations significantly impacted the average performance and variance across replicate trainings, while data ordering had little to no effect. Combining multiple augmentations produced the best performance on external validation (c-index +4.76%, HR +44.39%) compared to no augmentations, while most impactful individual augmentation type was additive gaussian noise (c-index +3.17%, HR +23.47%). Our results demonstrate that careful selection of strategies for data partitioning and ordering, weight initialization, and image augmentation are critical for developing robust and generalizable DL models for continuous survival prediction using medical imaging data.
17 Apr 2026
Read appraisal →Science advances
Enhancing biomedical optical volumetric imaging via self-supervised orthogonal learning
Optical volumetric imaging grapples with inherent noise problems arising from photon budget constraints, light scattering, and space-bandwidth product bottlenecks, all of which degrade structural fidelity and become even more pronounced than in planar imaging. While deep learning presents potential for denoising, supervised methods are hindered by the impracticality of paired datasets, and existing self-supervised approaches fail to fully exploit the intrinsic volumetric structural redundancy inherent to optical imaging. Here, we present a self-supervised volumetric biomedical imaging denoiser (VALID) that leverages intrinsic three-dimensional spatial coherence for highly efficient volumetric denoising through a self-supervised orthogonal learning framework. VALID demonstrates robust denoising performance across diverse imaging modalities, including two- and three-photon microscopy, light-field microscopy, and optical coherence tomography, substantially enhancing structural fidelity in deep-tissue, multimodal, and dynamic imaging scenarios. By combining computational efficiency with zero-shot adaptability, VALID establishes a transformative approach to volumetric image enhancement with structure-aware precision.
16 Apr 2026
Read appraisal →European journal of trauma and emergency surgery : official publication of the European Trauma Society
Comparison of CATCH, PECARN and CHALICE rules in children with minor head trauma
PURPOSE: The aim of this study is to prospectively compare the diagnostic performance of the PECARN, CATCH, and CHALICE clinical decision rules in pediatric patients presenting with minor head trauma, within a rural secondary-level emergency department characterized by limited resources, high patient volume, and restricted access to specialist physicians. METHODS: The study was designed as a prospective observational study. The PECARN, CATCH, and CHALICE clinical decision rules were applied to the patients included in the study, and the data were recorded. The data included in the study were statistically analyzed using the SPSS (Statistical Package for Social Sciences; SPSS Inc., Chicago, IL) 27 program. A value of p < 0.05 was considered statistically significant. RESULTS: A total of 1,032 patients were included in the study. The PECARN did not miss any of the 15 patients with pathological brain CT and had 100% sensitivity and 52.1% specificity. CATCH failed to detect pathological brain CT in 1 patient and had 93.3% sensitivity and 53.3% specificity. CHALICE failed to detect pathological brain CT in 2 patients, with 86.7% sensitivity and 87.8% specificity. Overall, PECARN was more successful in detecting pathological brain CT, but it also had a high false-positive brain CT rate. CONCLUSION: PECARN criteria are more sensitive than other criteria in predicting brain CT pathology, but they lead to a large number of brain CT scans. While the CHALICE criteria are beneficial in reducing unnecessary brain CT scans, they may miss positive cases.
16 Apr 2026
Read appraisal →The Journal of clinical investigation
Disrupted neurovascular-endocrine coupling in type 1 diabetes with impaired awareness of hypoglycemia
BACKGROUNDRecurrent hypoglycemia in type 1 diabetes (T1D) may culminate in impaired awareness of hypoglycemia (IAH). While neuroimaging studies identified affected brain regions, more complex perspectives integrating vascular dynamics with endocrine profile are needed.METHODSHere, 26 healthy adults, 30 T1D patients with normal hypoglycemia awareness (NAH), and 25 T1D patients with IAH underwent a hyperinsulinemic stepped clamp (euglycemia → hypoglycemia 50 mg/dL) combined with pseudo-continuous arterial spin-labeling MRI. Cerebral blood flow (CBF) and sympathetic vasomotor range (0.02-0.05 Hz) CBF oscillations were modeled against serially sampled plasma cortisol, epinephrine, norepinephrine, and glucagon.RESULTSIn healthy individuals treated as controls, hypoglycemia evoked robust thalamo-striatal and salience-interoceptive CBF increases (mean Cohen's d across significant clusters = 0.93) and suppression of vasomotor oscillations (d = 0.71). T1D retained CBF response but failed to attenuate oscillations (dT1D>controls = 0.43). IAH further blunted hypoglycemia-associated CBF increase, especially in thalamus, striatum, and insula (dNAH>IAH = 0.51). Hormone-CBF coupling differed quantitatively: cortisol/epinephrine-CBF correlations were positive in controls (r = 0.37/0.26), negative in NAH (-0.16/-0.40), and strongly positive in IAH (0.42/0.46).CONCLUSIONThus, our findings indicate that T1D disrupts dynamic, sympathetic modulation of CBF, whereas IAH additionally impairs perfusion reserve and shows maladaptive catecholamine-dependent CBF regulation, suggesting a qualitatively distinct neurovascular phenotype.TRIAL REGISTRATIONClinicalTrials.gov: NCT02747680 and NCT02866435.FUNDINGNIH (P41-EB-015894, P30-NS-076408, R01-DK-099137, R56-DK-099137, and DP1 AG093028); National Center for Advancing Translational Sciences of the NIH (KL2-TR-000113 and UL1-TR-000114); DP1 AG093028; Charles University, Czech Republic (Cooperatio Program, research area NEUR), Brain Dynamics (grant number CZ.02.01.01/00/22_008/0004643); General University Hospital in Prague (MH CZ-DRO-VFN64165).
16 Apr 2026
Read appraisal →Physics in medicine and biology
A deep learning-based framework for patient-specific radiation dose prediction in beta-emitting radionuclide therapies
Objective.Conventional image-based models for radionuclide therapy dosimetry are typically radionuclide-specific and rely on nuclear medicine (NM) images for training. We developed a deep learning (DL) model that predicts doses for not only training radionuclides but also radionuclides not included in the training data without patient NM images.Approach.The DL model was trained to predict voxel dose kernels (VDKs) for beta-emitters using CT-derived density maps and energy spectra. The DL model-derived VDKs were convolved with time-integrated activity to generate patient-specific dose maps. Zero-shot prediction was tested on Ho-166 and Lu-177. The performance of DL model was validated at kernel, phantom, and patient levels. The proposed DL model, conventional DL model trained on Y-90 NM/CT images of 22 patients, voxelS-value (VSV), and organ-level dosimetry (OLINDA/EXM) were benchmarked against Monte Carlo simulation (MC).Main results.Training utilized a single CT image, circumventing large patient cohort requirements. The DL model-based VDKs showed strong correlation with MC-based VDKs (R2: 0.99 and 0.91 for test Ho-166 and Lu-177) and, at the phantom level, achieved overall lower errors than VSV in heterogeneous region. At the patient level, the proposed DL model predicted organ doses with the smallest differences from MC while reducing computation time from 4.6 h to 2.6 min. For the test Lu-177, the proposed model exhibited good agreement with MC (2.1 Gy, 11.6%), while the conventional model trained using Y-90 showed large discrepancies (129.4 Gy, 657.4%). For the mean tumor dose averaged across the patient cohort for test Ho-166, the proposed DL model yielded the most accurate results with a difference of only 1.1 Gy (2.9%), outperforming multiple VSV (3.8 Gy, 7.2%), single VSV (3.8 Gy, 7.3%), and OLINDA/EXM (1.4 Gy, 4.5%).Significance.The proposed DL model enables dose prediction for radionuclides not included in the training data without NM images.
16 Apr 2026
Read appraisal →Journal of medical Internet research
Error Detection in Emergency Radiology Reports Using a Large Language Model: Multistage Evaluation Study
BACKGROUND: Emergency radiology requires highly accurate reporting under time constraints; yet, increasing workloads raise the risk of errors. While large language models (LLMs) show potential for proofreading in general radiology, their performance in emergency settings and non-English contexts remains unclear. OBJECTIVE: We aim to evaluate the performance of a domain-optimized LLM, DeepSeek-R1, for identifying errors in Chinese emergency radiology reports, with comparison against assessments by board-certified radiologists. METHODS: We compiled 7435 emergency reports (dataset 1; radiography, computed tomography, and magnetic resonance imaging) collected from November 2024 to April 2025. In stage 1, a total of 5 LLMs were evaluated using 200 reports. The best model, DeepSeek-R1, proceeded to stages 2 and 3, where 0-shot and few-shot learning were tested on a separate set (n=100). Model performance was compared against 12 radiologists. Stage 4 validated real-world utility on 800 verified reports. RESULTS: In subdataset 1, under stress-testing conditions, DeepSeek-R1 achieved a higher error detection rate in the few-shot setting than in the 0-shot setting (84.4% vs 60.9%, P=.003). Its performance exceeded that of radiology residents (84.4% vs 51.6% and 53.1%, respectively; both P<.05) and showed no statistically significant difference compared with senior radiologists and attending radiologists (84.4% vs 68.8%-93.8%, P=.26 to ≥.99). Compared with residents, DeepSeek-R1 detected more critical omissions (100% vs 25% and 50%; both P<.05) and other errors (92% vs 33% and 33%; both P=.02). In dataset 2, collected from independent institutions, DeepSeek-R1 achieved a detection rate of 95% under the few-shot setting. Reading time was shorter than that of human readers (92 vs 109 s). In real-world validation, DeepSeek-R1 identified 117 true reporting errors, yielding a positive predictive value of 56.5%. CONCLUSIONS: DeepSeek-R1 holds promise for improving quality control in emergency radiology reports. Its performance and efficiency support its use as an assistive proofreading tool in real-world radiology workflows.
15 Apr 2026
Read appraisal →Scientific reports
Region guided mask R-CNN with Haralick ResNet fusion for accurate coronary artery disease detection in computed tomography angiography images
Coronary Computed Tomography Angiography (CCTA) is a non-invasive imaging technique used to visualize the coronary arteries and diagnose coronary artery disease. It provides detailed 3D images which helps in identifying blockages, stenosis and plaques. However, CCTA images often suffer from noise, low contrast, and motion artifacts, which complicate accurate segmentation and analysis. The proposed work addresses the challenges in detecting and classifying coronary artery disease (CAD) from CCTA images. For accurate coronary artery segmentation, Region-Guided Mask R-CNN (RG-Mask R-CNN) combines region growing for initial region identification with Mask R-CNN for precise instance segmentation. This hybrid approach reduces the complexity of the structures in CCTA images and provides better segmentation accuracy. For feature extraction, Haralick-ResNet Fusion (HRF) captures the texture information using Haralick features and ResNet for high-level feature extraction to handle the difficulty of subtle variation differentiation in CCTA images. Finally, a Deep Convolutional Network (DeepConvNet) combines deep learning and convolutional layers to classify CAD efficiently based on the features extracted. This integrated framework offers an accurate solution (98.3%) for CAD detection from CCTA images by overcoming common issues such as noise and complex arterial structures, while improving diagnostic performance.
15 Apr 2026
Read appraisal →European radiology experimental
Impact of deep learning image reconstruction on ADC quantification and histogram metrics: a phantom study
OBJECTIVE: Recently, deep learning (DL)-based reconstruction methods have been introduced into clinical magnetic resonance imaging (MRI) systems to enhance image quality and reduce acquisition time. However, their effects on apparent diffusion coefficient (ADC) maps remain unclear. We investigated whether DL-based image reconstruction influences ADC quantification and histogram-based ADC metrics using a calibrated diffusion-weighted imaging (DWI) phantom. MATERIALS AND METHODS: A phantom containing vials with known ADC values was scanned on a 3-T system using full (fFOV) and reduced (rFOV) field-of-view DWI sequences. Each acquisition was performed using conventional (DL-OFF) and three DL-based strength levels (low, medium, high). Median ADC values were analyzed for repeatability (coefficient of variation (CV)) and accuracy. Histogram changes and first-order radiomic features were assessed using the Wasserstein distance, Friedman, and Wilcoxon tests. RESULTS: ADC estimates showed high repeatability (CV 0.1-1.2%) and good accuracy (deviation -2 to 7%) across all DL levels and sequences. DL reconstruction progressively reduced histogram dispersion, particularly in high-ADC vials. Wasserstein distances increased with DL strength, confirming a progressive effect on ADC value distributions, while median ADC values remained unchanged. Entropy and interquartile range decreased significantly (p < 0.001), whereas kurtosis and skewness increased, with differences showing less stable and sequence-dependent statistical significance. CONCLUSION: DL-based reconstruction maintained accurate and repeatable ADC quantification while reducing the dispersion of ADC values. The effect was more evident for high-ADC regions and the rFOV sequence, resulting in narrower distributions of ADC values. Further investigations comparing different DL-based solutions are warranted to assess the generalizability of these findings in clinical settings. RELEVANCE STATEMENT: Over the past decade, ADC histogram analysis has proven valuable for quantifying tumor heterogeneity, differentiating tumor grade, and evaluating early treatment response. Deep learning reconstruction narrows ADC distributions and reduces dispersion, supporting its potential in oncologic DWI, while highlighting the need for patient-based validation studies. KEY POINTS: DL reconstruction preserved ADC accuracy in both full FOV and reduced FOV DWI. ADC repeatability remained high across DL levels for both DWI sequences. Histogram dispersion progressively reduced across DL levels, particularly in high-ADC vials. Entropy and interquartile ranges decreased progressively with increasing DL strength.
14 Apr 2026
Read appraisal →Journal of medical Internet research
Context-Aware Sentence Classification of Radiology Reports Using Synthetic Data: Development and Validation Study
BACKGROUND: Automated structuring of radiology reports is essential for data utilization and the development of medical artificial intelligence models. However, manual annotation by experts is labor-intensive, and processing real clinical data through commercial large language models (LLMs) presents significant privacy risks. These challenges are particularly pronounced for non-English languages like Japanese, where specialized medical corpora are scarce. While synthetic data generation offers a potential privacy-preserving alternative, its effectiveness in capturing complex clinical nuances-such as negation and contextual dependencies-to train robust classification models without any real-world training data has not been fully established. OBJECTIVE: This study aimed to develop a context-aware sentence classification model for Japanese radiology reports using an entirely synthetic training pipeline, thereby eliminating reliance on real-world clinical data during the development phase. Furthermore, we sought to evaluate the generalizability of this approach by validating the model's performance on diverse, multi-institutional, real-world reports. METHODS: Japanese radiology reports (n=3104) were generated using GPT-4.1 and automatically annotated at the sentence level into 4 categories (background, positive finding, negative finding, and continuation) using GPT-4.1-mini. The synthetic data were partitioned into training (n=2670), validation (n=334), and test (n=100) sets. We fine-tuned several models, including lightweight local LLMs (Qwen3 and Llama 3.2 series) using low-rank adaptation and Japanese text classification models (Bidirectional Encoder Representations from Transformers [BERT]-base Japanese v3, Japanese Medical Robustly Optimized BERT Pretraining Approach [JMedRoBERTa]-base, and ModernBERT-Ja-130M). External validation was performed using 280 real-world reports (3477 sentences) from 7 institutions in the Japan Medical Image Database, with ground-truth labels established by board-certified radiologists. Evaluation metrics included accuracy, macro-averaged F1 (macro F1) score, and positive predictive value for positive findings (PPV_1). RESULTS: All models achieved high performance on the synthetic test set (accuracy: 0.938-0.951; macro F1-score: 0.924-0.940). Overall performance declined on the external validation dataset (accuracy: 0.783-0.813; macro F1-score: 0.761-0.790), reflecting distributional differences between synthetic and real-world reports; however, PPV_1 remained stable and high across datasets (eg, 0.957 on the synthetic test set vs 0.952 on the external validation dataset for Qwen3 [4B]). Parsing errors occurred in LLM-based approaches (19-260 sentences, 0.55%-7.48% in the external dataset). CONCLUSIONS: This study demonstrates the feasibility of developing context-aware sentence classification models for Japanese radiology reports using a training pipeline based entirely on synthetic data. The stability of PPV_1 indicates that the models successfully captured the essential clinical terminology and linguistic patterns required to identify positive findings in real-world reports, despite the observed performance degradation during external validation. This approach substantially reduces manual annotation requirements and privacy risks, providing a scalable foundation for constructing structured radiology datasets to support the development of clinically relevant medical artificial intelligence models.
11 Apr 2026
Read appraisal →PloS one
Voxel-wise deep learning segmentation of hydroxyapatite and iodine in spectral photon-counting CT: A quantitative phantom study
Accurate non-invasive identification of hydroxyapatite (HA) deposits is important for diagnosing calcific musculoskeletal disease and quantifying vascular calcification, but conventional and dual-energy CT often struggle to distinguish HA from iodinated contrast because of overlapping attenuation, noise, and beam-hardening artifacts. Spectral photon-counting CT (SPCCT) offers improved energy resolution and spatial fidelity, yet most deep-learning approaches in spectral CT focus on continuous density regression or anatomical segmentation rather than direct voxel-wise material labeling. We developed SPFF-UNet, a spectral-preserving 3D segmentation model for direct classification of HA and iodine concentrations from five-bin SPCCT volumes without material-decomposition preprocessing. A cylindrical phantom containing twelve materials was scanned at 0.1 mm isotropic resolution, including five HA concentrations, three iodine concentrations, three soft-tissue equivalents, and water. SPFF-UNet integrates spectral squeeze-excitation, EnergyFiLM, and FourierGate to preserve and exploit multi-energy information throughout the network. The model was trained for thirteen-class voxel-wise segmentation and compared with five established 3D architectures under matched training conditions. SPFF-UNet achieved the best macro-averaged performance on a held-out phantom scan (Dice 0.72 ± 0.01, IoU 0.59 ± 0.01, sensitivity 0.73 ± 0.01, precision 0.71 ± 0.01), outperforming the strongest comparator, ResUNet++ (Dice 0.66 ± 0.02, IoU 0.46 ± 0.02, sensitivity 0.67 ± 0.02, precision 0.61 ± 0.03). Performance gains were concentrated in mid/low-contrast HA and low-concentration iodine, with reduced slice-wise variability and fewer HA-iodine misclassifications. These results suggest that preserving spectral information and applying targeted spectral modulation can improve concentration-aware voxel classification from SPCCT. This phantom-based proof-of-concept provides a basis for future in vivo validation.
11 Apr 2026
Read appraisal →Neurology
Neurovascular Unit Dysfunction and Cerebral Small Vessel Disease in Parkinson Disease: A Systematic Review and a Meta Analysis
BACKGROUND AND OBJECTIVES: Parkinson disease (PD) pathogenesis remains incompletely understood; beyond nigrostriatal loss, nondopaminergic mechanisms including neurovascular unit dysfunction may contribute to disability. Cerebral small vessel disease (CSVD) burden reflects neurovascular dysfunction in the form of white matter hyperintensities (WMH), lacunes, cerebral microbleeds (CMB), and enlarged perivascular spaces (ePVS). In PD cohorts, CSVD burden correlates with worse motor and gait scores. The aim of the study was to explore whether patients with PD exhibit greater CSVD burden than healthy controls (HC). METHODS: We conducted a PRISMA-conformant systematic review and meta-analysis of studies including adults with idiopathic PD and a HC group that presented data comparing CSVD burden in these 2 groups. Six databases (MEDLINE, Embase, CINAHL Plus, CENTRAL, Scopus, and Web of Science) were searched on May 14 and 15, 2024. Two reviewers independently screened records, extracted data, and assessed risk of bias, with discrepancies resolved by consensus. Continuous outcomes were pooled as standardized mean differences (SMD); dichotomous outcomes as odds ratios (ORs). We evaluated small-study effects for pooled analyses with 10 or more studies using funnel plots and Egger regression test. RESULTS: We examined 13,403 records. Forty-six studies (45 cross-sectional) met inclusion criteria, totaling 3,817 PD and 2,593 HC (mean ages: 66.9 and 66.5, respectively). WMH volume (k = 21) was higher in PD (SMD 0.36, 95% CI 0.11-0.62). Visual ratings also indicated higher WMH in PD: global (k = 14) SMD 0.27 (95% CI 0.08-0.46); periventricular (k = 11) SMD 0.32 (95% CI 0.12-0.51); deep (k = 8) SMD 0.20 (95% CI 0.09-0.31). Differences in CMB (k = 6; OR 1.18, 95% CI 0.57-2.42) and lacunes (k = 4; OR 1.48, 95% CI 0.58-3.78) were not significant. ePVS results were heterogeneous but trended toward greater burden in PD, most notably in the midbrain (k = 3; SMD 1.80, 0.15-3.44). Overall evidence quality was rated as low, reflecting the observational nature of the included studies. DISCUSSION: Our analysis showed PD to be associated with greater WMH burden and increased midbrain ePVS. Pooled differences in CMB and lacunes were not significant. Substantial heterogeneity and cross-sectional designs limit certainty; standardized imaging and prospective cohorts are needed to define mechanisms and clinical implications.
10 Apr 2026
Read appraisal →Biomedical physics & engineering express
Efficient machine learning models leveraging DCE-MRI morphological and dynamic features allow accurate breast lesion classification.
We propose an ensemble learning approach to classify malignant versus benign breast lesions by leveraging morphological and dynamic features derived from Magnetic Resonance Images (MRI). The analysis was performed on 164 breast lesions of the publicly available 'Advanced MRI Breast Lesions' dataset from The Cancer Imaging Archive, containing T2-weighted and Dynamic Contrast-Enhanced (DCE)-MRI sequences, along with the segmentation masks of suspicious lesions. After extracting radiomic features using the Pyradiomics Python package, we computed dynamic features from DCE-MRI kinetic curves, which describe the contrast agent wash-in and wash-out. These features have been defined as the derivatives of image intensity measures, such as mean and standard deviation, computed within the masks on the five DCE-MRI time steps. We trained and evaluated an eXtreme Gradient Boosting (XGBoost) classifier, experimenting with different combinations of features and segmentation masks in a stratified 5-fold cross-validation. The best model trained on T2-weighted MRI morphological features achieved an Area Under the Curve (AUC) score of 0.83 ± 0.04 (95% C.I. 0.826-0.835) on the independent test set consisting of 20 lesions, while the model using only dynamic features performed an AUC of 0.91 ± 0.03 (95% C.I. 0.907-0.915). Despite being obtained on a small test sample, these results show the potential of features derived from DCE images for breast lesions classification.
10 Apr 2026
Read appraisal →Scientific reports
Research on lung nodule detection in X-ray plain films based on improved YOLOv12 model
To improve automatic lung nodule detection in chest X-ray images, this study proposes an improved YOLOv12-based detection framework by integrating space-to-depth convolution (SPDConv), a dynamic upsampling module (DySample), and a lightweight feature aggregation module (VoVGSCSP). SPDConv enhances spatial information preservation during feature extraction, DySample replaces conventional upsampling to improve multi-scale feature fusion, and VoVGSCSP strengthens feature representation while reducing computational redundancy. The optimized YOLOv12-DSV model was trained and evaluated using a publicly available chest X-ray dataset with annotated lung nodules from the Roboflow platform, with performance assessed through five-fold cross-validation and external testing. Experimental results show that the proposed model achieved an mAP50 of 0.735 and an mAP50-95 of 0.426, outperforming the original YOLOv12 model (mAP50: 0.704; mAP50-95: 0.411). In addition, the proposed model reduced the number of parameters from 2.52 to 2.21 M, decreased FLOPs from 6.0 to 5.2 G, and increased inference speed from 97.6 to 107.8 FPS. These results indicate that the proposed YOLOv12-DSV model improves detection accuracy while reducing computational cost, achieving a more favorable balance between detection performance and model complexity for lung nodule localization in chest X-ray images.
10 Apr 2026
Read appraisal →BMJ paediatrics open
Does osteogenesis imperfecta predispose infants to metaphyseal fractures? A systematic review.
OBJECTIVE: The notion that osteogenesis imperfecta (OI) predisposes children to metaphyseal fractures is controversial. This review aimed to ascertain the prevalence of metaphyseal fractures in children with OI less than 2 years of age. MATERIALS AND METHODS: Medline, CINAHL and Web of Science were searched for studies reporting metaphyseal fractures in children with OI less than 2 years of age. Articles not retrievable, not published in English, or where 'metaphyseal' fracture was not defined/illustrated, were excluded. Article quality was independently assessed using the Critical Appraisal Skills Programme and Case Reports tools. RESULTS: Of 298 retrieved abstracts, 82 were duplicates: 216 articles were screened by title and abstract, excluding 186. Hand search of the remaining 30 articles identified two further articles. Only four articles were included: (1) 4/41 children with OI aged between 1 day and 3 years with 'metaphyseal' fractures; (2) a case series including a 16-month-old boy with three 'metaphyseal' fractures. Abuse concerns remained despite a type IV OI diagnosis (not genetically confirmed); (3) a 7-month-old girl with two 'corner' fractures suspected to have OI and to have been abused; and (4) a pair of siblings with 'metaphyseal' fractures and type XI OI (FKBP10 mutation). CONCLUSIONS: There is limited evidence that OI predisposes infants and young children to metaphyseal fractures. The two diagnoses are not mutually exclusive and until further large studies are conducted, physical abuse should always be considered in infants and young children with metaphyseal fractures, even in those with confirmed OI.
9 Apr 2026
Read appraisal →Physics in medicine and biology
SSI-Net: a hybrid physics-constrained deep learning framework for quantitative ultrasound speed-of-sound reconstruction
Objective.Quantitative ultrasound tomography faces challenges in reconstructing speed-of-sound (SoS) distributions due to the ill-posed nature of the inverse problem and the computational complexity of full-waveform inversion. This study aims to develop an efficient and physically consistent deep learning framework for accurate SoS mapping from ultrasound signals.Approach. We propose speed of sound inversion network (SSI-Net), a dual data-physics-driven framework that integrates a bidirectional gated recurrent unit encoder for temporal feature extraction, a U-Net decoder for high-resolution spatial mapping, and a physics-constraint module based on an exact, differentiable finite-difference time-domain solver of the nonlinear Westervelt equation. The network model is trained through a joint optimization strategy that minimizes both data-fidelity loss and physics-informed residual loss, with a two-phase training schedule to ensure stable convergence.Main results.SSI-Net was validated across simulated, tissue-mimicking phantom, andin vivomouse datasets. It achieved superior reconstruction accuracy, with peak signal-to-noise ratio improvements of 7.4%-18.9% over state-of-the-art methods, while reducing the physics-based residual by 25.6%-57.5%. The framework demonstrated strong generalization to different transducer geometries (ring and linear arrays) and maintained stable performance across a ten-fold frequency range. Inference was completed within 109.4 ms per sample.Significance.By embedding an exact wave solver into a trainable architecture, SSI-Net combines the representational capacity of deep learning with strict physical consistency. It provides a robust, efficient, and clinically translatable solution for quantitative SoS imaging, offering a promising tool for tissue characterization in diagnostic ultrasound.
9 Apr 2026
Read appraisal →Translational vision science & technology
OptiDON-Guard: An Ensemble Model for Identification of Dysthyroid Optic Neuropathy Using Optical Coherence Tomography Angiography
PURPOSE: To explore the value of machine learning models based on optical coherence tomography angiography (OCTA) in identifying dysthyroid optic neuropathy (DON). METHODS: A retrospective, dual-center, cross-sectional study was conducted involving 346 thyroid eye disease patients (with/without DON) from two institutions. Participants underwent ophthalmic examinations, including OCTA scans centered on the macula and optic nerve head. The OCTA data from Center 1 were split into a training set and an internal validation set, while the data from Center 2 served as an external test set. Features were extracted from OCTA images, and a wrapper-based feature selection strategy identified the optimal subset. A two-tiered stacking modeling framework was used, incorporating baseline model construction, data augmentation using pseudo-labeling, and ensemble integration. Model performance (area under the curve [AUC], accuracy, sensitivity, specificity, precision) was evaluated alongside receiver operating characteristic, calibration, and decision curve analyses. Additionally, the SHapley Additive exPlanation method was used to interpret feature contributions. RESULTS: The ensemble model, OptiDON-Guard, demonstrated superior performance to individual models on both internal validation and external test sets. In the internal validation set, it achieved an AUC of 0.872, accuracy of 0.943, sensitivity of 0.667, specificity of 0.979, and precision of 0.943. In the external test set, it maintained strong performance with an AUC of 0.854, accuracy of 0.816, sensitivity of 0.767, specificity of 0.842, and precision of 0.854. CONCLUSIONS: The OptiDON-Guard is a reliable, interpretable, and clinically applicable tool for identifying DON. TRANSLATIONAL RELEVANCE: The OptiDON-Guard offers a valuable clinical tool to enhance the diagnostic precision of DON.
9 Apr 2026
Read appraisal →Journal of the American Heart Association
Effects of Computed Tomography Hyperdense Middle Cerebral Artery Sign on the Efficacy of Endovascular Therapy in Acute Ischemic Stroke With Large Infarcts: A Subgroup Analysis of the ANGEL-ASPECT Trial
BACKGROUND: The hyperdense middle cerebral artery sign (HMCAS) observed on noncontrast computed tomography is associated with thrombus composition and thrombectomy outcomes. The impact of HMCAS on the efficacy of endovascular therapy (EVT) in patients with large core infarcts remains unclear. METHODS: This analysis uses data from the ANGEL-ASPECT (Endovascular Therapy in Acute Anterior Circulation Large Vessel Occlusive Patients With a Large Infarct Core) trial, a multicenter randomized controlled trial conducted in China. Patients with acute ischemic stroke and anterior-circulation large-vessel occlusion were categorized according to whether HMCAS was present on baseline noncontrast computed tomography. The primary outcome was the 90-day modified Rankin Scale score. RESULTS: Of the 432 patients included in this analysis, 33% were HMCAS positive on baseline noncontrast computed tomography. In the EVT-treated patients, patients with HMCAS had worse functional outcomes than those without HMCAS (adjusted relative risk, 0.44 [95% CI, 0.26-0.73]; P=0.002). Patients with HMCAS required a greater number of thrombectomy passes (P<0.001). Among patients without HMCAS, EVT was associated with better functional outcomes than medical management (generalized odds ratio [OR], 2.78 [95% CI, 1.80-4.27]; P<0.001), whereas this benefit was not statistically significant among patients with HMCAS (adjusted OR, 1.69 [95% CI, 0.93-3.08]; P=0.09). There was no significant interaction between HMCAS status and treatment assignment (P=0.19). CONCLUSIONS: In this subgroup analysis comparing EVT with medical management, we found no statistically significant treatment-by-HMCAS status interaction, indicating that the benefits of EVT extend to patients with large core infarcts irrespective of HMCAS status. However, the magnitude of benefit appears greater in patients without HMCAS, whereas those with HMCAS tend to have a poorer overall prognosis. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique Identifier: NCT04551664.
9 Apr 2026
Read appraisal →PloS one
Hybrid deep learning and feature selection approach for autism detection from rs-fMRI data
Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that is primarily characterized by deficits in social communication and restricted or repetitive behavioral patterns. Although psychologists contribute significantly to the understanding of ASD, offering insights into its cognitive, emotional, and behavioral dimensions through assessments, diagnoses, therapeutic approaches, and family support, the diagnostic process remains complex. This complexity arises from the diverse manifestations of the disorder and the challenges associated with data sharing. In addition, conventional machine learning approaches for ASD detection may struggle with high-dimensional neuroimaging data and may require careful feature engineering. Consequently, this motivated us to enhance ASD diagnosis by incorporating deep learning (DL) techniques for feature extraction alongside a modified exponential-trigonometric optimization (ETO) algorithm as a feature selection (FS) technique. The modified ETO integrates the Arithmetic Optimization Algorithm (AOA) and the Guided Learning Strategy (GLS) to improve diagnostic performance. To evaluate the effectiveness of the proposed model, we utilized resting-state functional MRI (rs-fMRI) data from the Autism Brain Imaging Data Exchange (ABIDE I). Furthermore, the performance of the proposed model was compared with that of established models. The results indicate that the proposed model achieves competitive and, in most cases, superior performance compared with the benchmark methods, demonstrating superior accuracy, sensitivity, and AUC in diagnosing ASD. On average across the three atlas-based feature sets, the proposed model has an accuracy, sensitivity, and AUC of 73%, 78%, and 79%, respectively.
9 Apr 2026
Read appraisal →Journal of medical Internet research
Initial Insights Into an Institutional Secure Large Language Model for Magnetic Resonance Imaging Examination Requests: Retrospective Study
BACKGROUND: Incomplete clinical details on magnetic resonance imaging (MRI) examination requests (MERs) can lead to suboptimal protocol selection. An institutional secure large language model (sLLM) with access to manually retrieved salient data from the electronic medical record (EMR) may improve request completeness and protocol accuracy across multiple MRI subspecialties. OBJECTIVE: The objective of this study was to compare clinician MERs with sLLM-augmented MERs for information quality and to evaluate the protocoling accuracy of the sLLM versus board-certified radiologists across body, musculoskeletal, and neuroradiology MRI. METHODS: This retrospective study included 608 random outpatient MRI examinations performed between September 2023 and July 2024 (body 206, musculoskeletal 203, neuroradiology 199). The cohort comprised 528 patients (mean 51.2 years, SD 19.2; range 4-93; n=279, 52.8% women, n=249, 47.2% men). MERs without EMR access were excluded. A privately hosted Anthropic Claude 3.5 model (temperature 0) augmented each MER with manually retrieved salient EMR data and, via rule-based parsing, mapped the extracted elements onto predefined institutional criteria to recommend region or coverage and contrast use. Two experienced radiologists established a consensus reference standard. Two board-certified general radiologists (Rad 3 and Rad 4) and the sLLM were compared with this standard. Clinical information quality was graded using the Reason-for-Exam Imaging Reporting and Data System (RI-RADS). Interrater reliability was quantified with Gwet AC1. Paired accuracies were compared with the McNemar test to determine whether there was a statistically significant difference. RESULTS: Interreader agreement for RI-RADS was almost perfect for sLLM-augmented MERs (AC1 0.97, 95% CI 0.94-0.99) and moderate for clinician MERs (AC1 0.43, 95% CI 0.34-0.52). Limited or deficient clinical information (RI-RADS C/D) fell to 0% to 0.7% (0/608 to 4/608) with sLLM augmentation vs 4.1% to 20.4% (25/608 to 124/608) for clinician MERs. Overall protocol accuracy was 93.1% (566/608; 95% CI 89.6-96.6) for the sLLM, 91.4% (556/608; 95% CI 87.6-95.3) for Rad 3, and 92.1% (560/608; 95% CI 88.4-95.8) for Rad 4 (sLLM vs Rad 3 P=.23 vs Rad 4 P=.40). Region or coverage accuracy was similar (sLLM: 579/608, 95.2%; Rad 3: 585/608, 96.2%; Rad 4: 573/608, 94.2%; P=.46 and P=.36). Contrast decisions were more accurate using the sLLM at 94.4% (574/608; 95% CI 91.3-97.5) vs Rad 3 at 92.1% (560/608; 95% CI 88.4-95.8; P=.027) and were not significantly different to Rad 4 at 92.9% (565/608; 95% CI 89.4-96.4; P=.16). Subspecialty analyses showed similar patterns, with the sLLM outperforming Rad 4 for musculoskeletal MRI contrast decisions (96.6% vs 91.1%; P=.006) and matching readers elsewhere. Manual review indicated that sLLM improvements arose from EMR details not listed on the MER (infection/inflammation, tumor history, prior surgery). No clinically significant hallucinations were identified in a manual review of discordant cases. CONCLUSIONS: Across body, musculoskeletal, and neuroradiology MRI, sLLM-augmented examination requests improved clinical context and enhanced contrast selection while demonstrating accuracy comparable to general radiologists for region or coverage. Integrating sLLMs into routine vetting workflows may reduce manual workload in protocol selection for more efficient, standardized protocoling.
8 Apr 2026
Read appraisal →The Cochrane database of systematic reviews
Antibiotic treatment for low back pain, radicular pain, or both.
RATIONALE: Low back pain is the leading cause of disability worldwide. Research has associated Modic changes (signal changes in the vertebral endplate on magnetic resonance imaging (MRI)) with low back pain. Some hypothesise that Modic type 1 changes (oedema) may have an infective cause, and that antibiotics could serve as targeted treatment. OBJECTIVES: To assess the benefits and harms of antibiotic therapy, compared with placebo or another treatment, in people with low back pain, radicular pain, or both. SEARCH METHODS: We searched CENTRAL, MEDLINE, Embase, and two trials registries up to 26 August 2025, with no restrictions related to language or date of publication. ELIGIBILITY CRITERIA: We included randomised controlled trials (RCTs) that assessed antibiotics versus placebo (primary comparison) or other treatments. We assessed the evidence separately for four populations of adults with low back pain, radicular pain, or both: those with 1) Modic type 1 changes and evidence of disc herniation, 2) other Modic changes and evidence of disc herniation, 3) any Modic changes but no evidence of disc herniation, and 4) evidence of disc herniation without any Modic changes. OUTCOMES: Critical outcomes were pain intensity and disability or function up to 12 to 14 weeks, and the number of participants with any adverse events and serious adverse events at the end of the trial period. RISK OF BIAS: Two review authors independently evaluated the risk of bias using the Cochrane risk of bias tool RoB 1. SYNTHESIS METHODS: Two review authors independently screened studies and extracted data. We synthesised results for each outcome using random-effects meta-analysis. Where this was not possible due to the nature of the data, we reported the results descriptively. We used GRADE to assess the certainty of evidence. The primary comparison was antibiotics versus placebo in people with low back pain, radicular pain, or both, and Modic type 1 changes with evidence of disc herniation (population 1) up to 12 to 14 weeks. INCLUDED STUDIES: We included three trials, conducted in Denmark, Belgium, and Norway, with a total of 402 participants (mean ages ranged from 44.7 years to 51.0 years). Two trials recruited people with chronic (> 6 months) moderate-to-severe low back pain (with or without radicular pain), Modic changes, and disc herniation. One trial included people with Modic type 1 changes only (population 1) and compared oral amoxicillin-clavulanate 500 mg/125 mg to placebo (for both treatments, 1 or 2 tablets 3 times daily for 100 days). The second trial included people with either Modic type 1 changes (population 1) or Modic type 2 changes (population 2) and compared oral amoxicillin 750 mg to placebo (for both treatments, 3 times daily for 100 days). Follow-up was 12 months in both trials. The third trial recruited people with lumbar radicular pain (with or without low back pain) and disc herniation. Modic changes was not an inclusion criterion (population 4). This trial evaluated minocycline 100 mg versus amitriptyline 25 mg versus placebo and followed participants for 14 days (treatment duration). There were no trials with population 3. SYNTHESIS OF RESULTS: All three trials were at low risk of selection, performance, and detection bias. We rated one trial at high risk of attrition bias and unclear risk of reporting and other bias. The overall certainty of evidence is low to very low, primarily due to imprecision and indirectness. In people with low back pain, Modic type 1 changes, and disc herniation, the antibiotic oral amoxicillin (with or without clavulanate) may improve pain slightly and may result in a small to moderate improvement in disability at 12 to 14 weeks compared with placebo (low-certainty evidence, downgraded for imprecision and indirectness). At this time point, the mean back pain score was 50.6/100 (where a lower score indicates less pain) in the amoxicillin group versus 59/100 in the placebo group (mean difference (MD) -8.42 points, 95% CI -16.18 to -0.67; I² = 35%; 2 trials, 255 participants), and the mean disability score was 45.2/100 (where a lower score indicates less disability) in the amoxicillin group versus 55.7/100 in the placebo group (MD -10.52 points, 95% CI -15.94 to -5.09; I² = 0%; 2 trials, 255 participants). We are unsure if amoxicillin (with or without clavulanate) results in more people reporting adverse events (very low-certainty evidence, downgraded for imprecision, inconsistency, and indirectness). Adverse events were reported in 76 of every 100 people taking amoxicillin compared with 49 of every 100 people taking placebo (risk ratio (RR) 1.79, 95% CI 0.54 to 5.94; I² = 96%; 2 trials, 262 participants). We are unsure if amoxicillin (with or without clavulanate) results in more people reporting serious adverse events (very low-certainty evidence, downgraded for imprecision and indirectness). Serious adverse events were reported in three of every 100 people taking amoxicillin (with or without clavulanate) compared with two of every 100 people taking placebo (RR 1.43, 95% CI 0.11 to 18.35; I² = 45%; 2 trials, 262 participants). We are unsure of the effects of amoxicillin compared to placebo in people with low back pain, Modic type 2 change, and disc herniation. Similarly, we are unsure of the effects of minocycline compared to placebo or amitriptyline in people with low back pain, disc herniation, and no Modic changes. AUTHORS' CONCLUSIONS: In people with low back pain, Modic type 1 changes, and evidence of disc herniation, the antibiotic amoxicillin (with or without clavulanate) may provide slight to small benefits in reducing back pain and small to moderate benefits in improving disability compared to placebo at 12 to 14 weeks. The evidence on the risk of adverse events, including serious adverse events, with amoxicillin (with or without clavulanate) is very uncertain. Further research is likely to change our confidence in the estimates. FUNDING: This review had no dedicated funding. REGISTRATION: Protocol (2021): DOI 10.1002/14651858.CD014221.
8 Apr 2026
Read appraisal →PloS one
A multi-modal data fusion and real-time monitoring on stroke risk prediction using federated learning
Predicting the risk of stroke is one of the critical problems in healthcare, which necessitates efficient solutions for providing accurate and prompt risk assessments while preserving data confidentiality. This work proposes a new framework using Federated Learning (FL) to combine Multi-Layer Perceptron (MLP) and Gated Recurrent Unit (GRU) models that are essential in analyzing multimodal data. Implemented in Python, the approach incorporates two datasets: Dataset 1, which consists of Demographic data medical history, and lifestyle data, and the second dataset, which includes the normal condition and the affected stroke condition CT scan images. Imputation of missing values, feature normalization by Min-Max scaling, and handling of imbalanced classes with SMOTE make the data pre-processing procedures exhaustive. In FL architecture three clients -Client A, Client B, and Client C - process a split multimodal dataset containing static and sequential information. Each client independently trains an MLP-GRU model. Each is applied with MLP handling static features from Dataset 1 and GRU handling sequential features from Dataset 2. To update models, Federated Averaging is used on a central server, to create a global model that is then returned to the clients for further refinement. The accuracy of the proposed method averages 99.00% and surpasses other models by 2.5% including CNN, LSTM, Random Forest, and SVM. By enhancing MLP with GRU and applying them to a privacy-preserving FL framework,The study addresses the fragmented use of multimodal medical data, where clinical records and imaging are generally evaluated separately, resulting in inadequate diagnostic support. The strategy integrates complementary modalities to create a more comprehensive perspective of patient health, enhancing healthcare predictive accuracy and decision-making. This incentive is essential for improving computational methods and linking technical advancement with medical objectives like fast diagnosis and therapy planning. The introduction emphasises the therapeutic necessity of harmonising organized and unstructured data to reduce diagnostic ambiguity. A translational approach is used to discuss how multimodal integration might improve clinical workflows, develop collaborative healthcare systems, and support sustainable medical practices. This repeated emphasis links methodological advances to real-world healthcare issues, boosting the study's academic relevance sets.
7 Apr 2026
Read appraisal →Circulation
Wearable-Derived Training Load and Coronary Atherosclerosis in Middle-Aged and Older Athletes and Physically Active Controls: A New Perspective From the Master@Heart Study
BACKGROUND: Middle-aged and older endurance athletes have increased prevalence of coronary artery disease (CAD) on coronary computed tomography angiography compared with healthy controls, despite similarly low cardiovascular risk. Previous studies relied on self-reported data to quantify training load (TL), which poorly correlates with objective wearable-derived TL and may bias outcomes. The effect of objective TL on CAD risk remains unknown. METHODS: In this observational, cross-sectional analysis of the Master@Heart study, 222 men (median age, 54 [49-59] years) were included: 77 lifelong athletes, 98 late-onset athletes, and 47 controls. TL was assessed using objective wearable-derived training duration and intensity (12 consecutive months), as well as self-reported training measures. Coronary computed tomography angiography-derived CAD prevalence was compared across TL quartiles (Q) using a global unadjusted chi-square test and logistic regression, adjusted for cardiovascular risk factors and years of endurance exercise, to estimate odds ratios between Q4 and Q1. In addition, adjusted logistic regression models were fitted with continuous TL, using smoothing splines to capture potential nonlinear associations. RESULTS: Across quartiles of objective Edwards training impulse (training duration × heart rate-weighted intensity), unadjusted global differences were observed for ≥1 plaque (P<0.001), coronary artery calcification (CAC)>0 (P=0.002), and CAC>100 (P=0.012). Q4 participants had significantly higher adjusted odds of ≥1 plaque (odds ratio, 5.85; 95% CI, 2.33-14.71), CAC>0 (odds ratio, 5.03; 95% CI, 2.04-12.35), and CAC>100 (odds ratio, 3.50; 95% CI, 1.22-10.00) versus Q1. Similar associations were found for objective training duration, whereas no clear associations were observed for relative time spent in high-intensity zones. In continuous analyses, Edwards training impulse and objective training duration showed significant positive associations with ≥1 plaque and CAC>100 (P<0.05), whereas self-reported training duration was only significantly associated with CAC>100 (P<0.05). Metabolic equivalent of task-minutes per week based on self-reported TL was not associated with CAD (P>0.05). CONCLUSIONS: High training duration (hours/week), particularly when combined with cumulative high-intensity TL (Edwards training impulse), was independently associated with increased prevalence of subclinical CAD in middle-aged and older athletes and physically active controls. Exercise intensity alone, in the absence of high duration, was not clearly linked to CAD. These findings underscore the potential of objectively measured TL for understanding associations with subclinical CAD in endurance athletes. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique identifier: NCT03711539.
7 Apr 2026
Read appraisal →JMIR medical informatics
Deep Learning for Content-Based Medical Image Retrieval in Picture Archiving and Communication Systems for Brain Tumor Detection: Algorithm Development and Validation
BACKGROUND: Advances in medical imaging have led to massive archives, yet navigating these datasets remains challenging due to the limitations of traditional text-based search engines. While content-based medical image retrieval (CBMIR) offers a visual feature-based solution to enhance clinical workflows and research, its operational integration into picture archiving and communication systems (PACS) remains a significant bottleneck. Despite the progress in deep learning for feature extraction, CBMIR tools are rarely integrated and effectively implemented within existing radiology information systems due to complex protocol barriers. OBJECTIVE: To address these challenges, this study develops a CBMIR system meticulously designed to cater to 7 distinct types of brain tumors as seen in brain magnetic resonance images. Our system is tailored to assist radiologists and health care professionals in efficiently retrieving pertinent historical medical images, thereby providing quantitative decision support for radiologists and facilitating evidence-based case comparison, with the potential to improve retrieval efficiency and clinical workflow, rather than directly enhancing diagnostic accuracy. METHODS: The dataset used in this study was collected from a single medical center and is not publicly available. The core innovation is a state-of-the-art deep learning-based feature extraction algorithm specifically engineered for the CBMIR system. We use GoogLeNet as the primary architecture, incorporating generalized mean pooling to capture nuanced local features and an embedding layer for dimension reduction. Crucially, we address the integration gap by harmonizing 2 open-source projects to successfully embed the CBMIR system into a functional PACS environment via standard protocols. RESULTS: The image dataset contains 658 participants with 15,873 images collected from 2000 to 2017. The empirical findings of our research demonstrate the performance and robustness of the proposed CBMIR system. Our system achieves a remarkable mean average precision score of 89.16% and an equally impressive Precision@10 score of 94.08%. These metrics affirm the system's efficacy in retrieving relevant medical images. Furthermore, we successfully integrate the CBMIR system into a PACS by successfully harmonizing 2 open-source projects. CONCLUSIONS: This study presents the design and implementation of a PACS-integrated CBMIR system for brain magnetic resonance imaging, and the experimental results demonstrate that the system can achieve efficient and accurate image retrieval within a clinical workflow.
7 Apr 2026
Read appraisal →Translational vision science & technology
Diagnosis of High Intracranial Pressure by Non-Optic Nerve Retinal Image Features
PURPOSE: To assess the performance of an artificial intelligence deep learning (DL) model compared with neuro-ophthalmologists for the classification of subjects with elevated intracranial pressure (ICP) and papilledema versus control subjects, based on retinal images with and without masking the optic nerve heads (ONHs). METHODS: Widefield retinal images were obtained in 32 subjects (70 images) with elevated ICP and papilledema and 31 control subjects (62 images). ONH-unmasked images were generated by cropping the image to a 30° circular region. ONH-masked images were generated by masking the ONH and peripapillary region. Classification was performed using a convolutional neural network model and by two neuro-ophthalmologists (graders). RESULTS: For the ONH-unmasked images, the classification accuracies of the model, grader 1, and grader 2 were 83% (area under the receiver operating characteristic curve [AUC] = 0.94), 93%, and 86%, respectively. For the ONH-masked images, the classification accuracies of the model, grader 1, and grader 2 were 79% (AUC = 0.90), 66%, and 76%, respectively. The classification performance of the DL model and graders did not significantly differ for both datasets (P ≥ 0.38). There was no significant effect of ONH masking on the performance of the DL model (P = 1.0) and grader 2 (P = 0.38), whereas the performance of grader 1 was significantly reduced (P = 0.02). CONCLUSIONS: Both expert graders and the DL models demonstrated excellent performance for classifying retinal images of subjects with elevated and normal ICP using images with or without masking the ONH. TRANSLATIONAL RELEVANCE: Methods for assessment of non-optic nerve retinal image features have the potential to improve diagnosis and monitoring the progression and response to treatment of elevated ICP.
7 Apr 2026
Read appraisal →PloS one
A robust deep learning approach for impulse noise filtering using hybrid auto-encoder with fuzzy median filter
De-noising convolutional neural networks (DnCNNs), are a powerful nonlinear mapping models in image processing for impulse noise removal. During training and validation, a set of 12 standard testing images is used to evaluate model performance. DnCNNs demonstrate strong capability in classification of impulse noise with excellent results. To evaluate de-noising performance, a suitable noise ratio should be added so that most appropriate DnCNN model can be used for impulse noise detection. This research proposes an effective image restoration technique that integrates DnCNN and an autoencoder with a fuzzy median filter to detect and eliminate high-density impulse noise. The proposed deep learning de-noising technique used to classify noisy and clean pixels, and result are presented in different metrics such as accuracy, FPR, FNR and f1 score. Further, to remove impulse noise an auto-encoder with fuzzy median filter are used that then reconstructs the clean image based with parametric values. Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), are used to assess our methodology, it is compared to conventional impulse noise filtering techniques, experimental results indicate a significant improvement in image quality. Based on the final de-noised images, this research contributes to developing deep learning-based, de-noising techniques that enhance image restoration quality while preserving image details and essential features.
5 Apr 2026
Read appraisal →PloS one
Utilizing the transformer mechanism to predict cervical lymph node metastasis in patients with papillary thyroid carcinoma
RATIONALE AND OBJECTIVE: The status of cervical lymph node metastasis(LNM) in Papillary thyroid carcinoma(PTC) can affect the patient's treatment plan and prognosis. This study aims to develop and validate the application value of Vision transformer (ViT) model in preoperatively predicting cervical LNM in PTC. MATERIALS AND METHODS: A total of 540 PTC patients were retrospectively reviewed from two hospitals from April 20,2022 to August 20,2023.The ViT model is built based on the two-dimensional rectangular ultrasound image of the primary thyroid tumor, and at the same time, to compare its performance, a deep learning model of the traditional Convolutional neural network (CNN) framework, a ultrasound radiomics combined model(Clinical-Rad model), and clinical model are built. RESULTS: The ViT model demonstrated an AUC of 0.807 (95% CI: 0.709-0.905) in the internal validation cohort and 0.809 (95% CI: 0.720-0.900) in the external validation cohort. The ViT model's AUC ranged from 0.807-0.814 across all cohorts, significantly exceeding the clinical model (AUC: 0.595-0.669, P<0.001). While the AUC of the ViT model in the training cohort was slightly lower than that of the combined ultrasound radiomics model (0.814 vs 0.828, P=0.491), it showed significantly higher AUC values in the internal (0.807 vs 0.718, P=0.049) and external validation cohorts (0.809 vs 0.691, P<0.001). Compared to the clinical and combined radiomics models, the ViT model exhibited stable and superior predictive performance for PTC cervical lymph node metastasis.In the internal validation cohort, Doctor C's net reclassification improvement (NRI) with the ViT model was 0.106 (P=0.022), and the integrated discrimination improvement (IDI) was 0.106 (P=0.023). Doctor D showed NRI and IDI values of 0.113 (P=0.022) and 0.106 (P=0.024), respectively. In the external validation cohort, Doctor C's NRI and IDI were 0.090 (P=0.024) and 0.106 (P=0.024), while Doctor D had values of 0.011 (P=0.013) and 0.106 (P=0.013). The ViT model enhanced the diagnostic capabilities of both Doctor C, with less clinical experience, and Doctor D, with extensive experience. CONCLUSION: The deep learning model based on the Transformer mechanism shows good performance in predicting LNM in PTC patients, which is superior to the clinical model, Clinical-Rad model, and traditional CNN model.
5 Apr 2026
Read appraisal →AJNR. American journal of neuroradiology
Flow Diversion versus Coiling with or without Stenting for Unruptured Wide-Neck Intracranial Aneurysms: A Randomized Comparison
BACKGROUND AND PURPOSE: There are few randomized trials comparing coiling (with or without stent placement) and flow diversion (FD) for the treatment of wide-neck unruptured intracranial aneurysms. MATERIALS AND METHODS: EVIDENCE was an investigator-led randomized (1:1) trial conducted in 7 French university hospitals. Patients with 7 to 20-mm intradural unruptured intracranial aneurysms with a 4- to 10-mm neck and a "dome/neck" ratio of ≥1 were randomly allocated to coiling with or without adjunctive stent placement or FD alone. The composite primary efficacy outcome was "treatment failure," defined as initial failure to treat the aneurysm; aneurysm rupture or retreatment during follow-up; death or dependency (mRS> 2); or an angiographic residual aneurysm adjudicated by an independent core laboratory at 12 months. The primary hypothesis (revised for slow accrual) was that FD would decrease treatment failures from 35% to 10%, requiring 90 patients. Primary analyses were intent to treat. RESULTS: Among the 91 enrolled patients, four (two in each group) withdrew consent; 87 patients were included in the analysis: 43 in the FD group and 44 in the control group. Most patients had <10 mm (57/87; 65.5%) asymptomatic ophthalmic aneurysms (75/87, 86.2%). A poor primary outcome, ascertainable in 86 patients, was reached in 8/43 patients with FD (18.6%; 95% CI, 9.7%-32.6%) compared with 10/43 coiling patients (23.3%; 95 CI, 13.1%-37.7%) (RR = 0.80; 95% CI, 0.35-1.83; P = .60). Serious adverse events were similar. CONCLUSIONS: For patients with mostly unruptured, wide-neck ophthalmic aneurysms of <10 mm, FD was not superior to coiling with or without stent placement.
5 Apr 2026
Read appraisal →BMJ open
Utilisation of brain MRI for the diagnosis of dementia in low and middle-income countries: systematic review and meta-analysis
OBJECTIVES: To synthesise the prevalence and patterns of dementia-relevant structural brain MRI abnormalities in adults with suspected or confirmed dementia in low- and middle-income countries (LMICs), and to summarise MRI protocols and the incremental diagnostic contribution of MRI beyond cognitive screening. DESIGN: Systematic review and meta-analysis. DATA SOURCES: PubMed, EMBASE, Web of Science and PsycINFO (January 1990-27 January 2025), plus reference list screening and targeted manual searches. ELIGIBILITY CRITERIA: Observational or diagnostic-accuracy studies from World Bank-defined LMICs including adults (≥50 years) with suspected or confirmed dementia who underwent brain MRI as part of diagnostic evaluation. DATA EXTRACTION AND SYNTHESIS: Two reviewers independently screened, extracted data and assessed risk of bias using ROBINS-I. Random-effects models pooled prevalence of dementia-relevant MRI abnormalities; diagnostic-accuracy outcomes were synthesised narratively due to heterogeneous reference standards and incomplete reporting. RESULTS: 39 LMIC studies were included; 23 studies (2513 participants) contributed to the meta-analysis. Dementia-relevant MRI abnormalities (defined as ≥1 clinically relevant structural abnormality per study definition) were present in 1248/2513 participants. The pooled prevalence of dementia-relevant MRI abnormalities was 58% (95% CI 43% to 72%), with substantial heterogeneity (I²=95%) and a wide prediction interval (8-96%), indicating marked between-study variability; this estimate should be interpreted as a descriptive summary of study-level proportions rather than a precise population parameter. CONCLUSIONS: Brain MRI frequently demonstrates dementia-relevant pathology in LMIC clinical cohorts, usually with mixed neurodegenerative-vascular patterns. Structured visual ratings may add aetiologic specificity beyond cognitive screening, but pooled estimates should be interpreted as summaries of heterogeneous study-level findings rather than precise population parameters, given high heterogeneity and risk of bias. PROSPERO REGISTRATION NUMBER: CRD42024510241.
4 Apr 2026
Read appraisal →Yonsei medical journal
Deep Learning for Automated Tumor Segmentation of Rectal Cancer on T2-Weighted Magnetic Resonance Images
PURPOSE: To develop a deep learning (DL)-based automated segmentation model for rectal cancer on T2-weighted (T2W) magnetic resonance (MR) images. MATERIALS AND METHODS: A total of 458 patients who underwent baseline rectal MR imaging were retrospectively enrolled. An experienced radiologist labeled the tumor on each slice of the selected T2W axial images covering the entire tumor mass, and tumor volume was measured during labeling. Another radiologist labeled the rectum on the same images. Attention U-Net was trained on the T2W images from the training dataset to classify each voxel as tumor or non-tumor. Segmentation with or without rectum guidance was compared using evaluation metrics, including the Dice similarity coefficient (DSC), in the test dataset. RESULTS: The tumor segmentation DSC without rectum guidance was 73.35% (71.97-74.71); that with the predicted rectum was 71.79% (70.53-73.05); and that with the overlaid rectum was 75.52% (74.32-76.62), which was the highest among the three segmentation models. Tumor volume showed a positive correlation with tumor segmentation accuracy with the predicted rectum (r=0.333, p=0.001; overlaid rectum, r=0.319, p=0.002) and without rectum guidance (r=0.219, p=0.036). CONCLUSION: A DL-based automated segmentation model can predict RC with over 70% accuracy, which was further improved with overlaid rectum, and larger tumor volume.
4 Apr 2026
Read appraisal →European radiology
Predictive value of artificial intelligence-based quantitative CT feature analysis for diagnosing the pathological types of pulmonary nodules
OBJECTIVES: Accurate preoperative classification of pulmonary nodules (PNs) is critical for guiding clinical decision-making and preventing overtreatment. This study aims to evaluate the predictive performance of artificial intelligence (AI)-based quantitative computed tomography (CT) feature analysis in differentiating among four pathological types of PNs: atypical adenomatous hyperplasia and adenocarcinoma in situ (AAH + AIS), minimally invasive adenocarcinoma (MIA), invasive adenocarcinoma (IAC), and lung inflammatory nodules (IN). MATERIALS AND METHODS: A total of 462 pathologically confirmed PNs were analyzed. Radiomic features, including CT attenuation metrics, 3D morphometrics, and texture parameters such as entropy and skewness, were extracted using a deep learning-based AI platform. Logistic regression models were constructed using both single- and multi-variable strategies to evaluate the classification accuracy of these features. Moreover, the inclusion of IN as a separate category significantly enhanced the clinical utility of AI in differentiating benign mimickers from malignant nodules. The combined model, which integrated AI-derived features with traditional CT signs, was used to assess the diagnostic performance of the radiomic features in differentiating the four pathological types of nodules. RESULTS: The combined model demonstrated superior diagnostic performance, with area under the curve (AUC) values of 0.936 for IAC, 0.884 for AAH + AIS, and 0.865 for IN. Although MIA showed lower classification accuracy (AUC = 0.707), key features such as entropy, solid component ratio, and total volume effectively distinguished invasive from non-invasive lesions. CONCLUSION: These findings highlight the potential of AI-enhanced radiomics for supporting non-invasive, objective, and individualized diagnosis of PNs. KEY POINTS: Question Can artificial intelligence (AI)-based quantitative CT analysis reliably differentiate benign inflammatory nodules from the spectrum of lung adenocarcinoma subtypes, a common diagnostic challenge? Findings An integrated model combining AI-driven radiomic features and traditional CT signs demonstrated high accuracy in differentiating invasive adenocarcinoma (AUC = 0.936), pre-invasive lesions (AUC = 0.884), and inflammatory nodules (AUC = 0.865). Clinical relevance AI-enhanced radiomics provides a non-invasive, objective tool to improve preoperative risk stratification of pulmonary nodules, potentially guiding personalized management and reducing unnecessary surgeries for benign inflammatory lesions that mimic malignancy.
4 Apr 2026
Read appraisal →Neurological research
Machine learning-based preoperative prediction of IDH mutation status in adult gliomas using clinical features
BACKGROUND: As isocitrate dehydrogenase (IDH) mutation status represents a critical prognostic factor in adult gliomas, there is a demand for a straightforward but effective predictive model to facilitate rapid preoperative diagnosis. METHODS: A cohort of 418 adult glioma patients from our institution were retrospectively enrolled and randomly partitioned into training (70%) and internal validation (30%) sets. Among 9 machine learning algorithms evaluated, AdaBoost demonstrated superior predictive performance based on receiver operating characteristic (ROC) analysis and was subsequently selected as the optimal model. Through feature importance ranking derived from the model, the top 8 most contributory features were identified and incorporated into the final streamlined prediction model. The model's generalizability was further assessed using an independent external validation cohort comprising 206 adult glioma cases. RESULTS: The AdaBoost model demonstrated robust discriminative performance, achieving an area under the receiver operating characteristic curve (AUC-ROC) of 0.901 in the internal validation cohort and 0.849 in the independent external validation cohort. SHAP (SHapley Additive exPlanations) interpretation revealed age as the most influential predictive feature, contributing significantly to the model's decision-making process. CONCLUSIONS: This validated AdaBoost model offers a practical and interpretable solution for preoperative IDH status assessment, demonstrating both diagnostic reliability and biological plausibility through its feature importance profile.
4 Apr 2026
Read appraisal →Computer methods in biomechanics and biomedical engineering
Reconstruction of blood flow velocity with deep learning information fusion from spectral ct projections and vessel geometry
In this work, we investigate a new deep learning reconstruction method of blood flow velocity within deformed vessels from contrast enhanced X-ray projections and vessel geometry. The principle of the method is to perform linear or nonlinear dimension reductions on the Radon projections and on the mesh of the vessel. These low dimensional projections are then fused to obtain the velocity field in the vessel. The accuracy of the reconstruction method is proved using various neural network architectures with realistic unsteady blood flows. The approach leverages the vessel geometry information and outperforms the simple PCA-net.
4 Apr 2026
Read appraisal →Japanese journal of radiology
Super-resolution deep learning reconstruction improves brain MRI quality and detection of metastases.
PURPOSE: Accurate identification of brain metastases is critical for determining prognosis and guiding treatment. Deep learning reconstruction (DLR) enhances MRI quality by reducing noise, while super-resolution DLR (SR-DLR) may further improve spatial resolution and lesion detectability. To evaluate SR-DLR versus conventional DLR in detecting and visualizing brain metastases on postcontrast T1-weighted brain MRI. MATERIALS AND METHODS: This retrospective study included 47 consecutive patients who underwent postcontrast 3D whole-brain T1-weighted MRI between July and December 2024. Images were reconstructed using both SR-DLR and DLR. Three independent readers evaluated metastatic lesion detection and rated image quality. Subjective assessments included lesion visibility, visibility of normal structures, sharpness, noise, and overall image quality. Objective metrics-full width at half maximum (FWHM), edge rise distance (ERD), edge rise slope (ERS), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR)-were also measured. Statistical tests included jackknife alternative free-response receiver operating characteristic (JAFROC) analysis, Wilcoxon signed-rank test, McNemar's test, and paired t-tests, with significance threshold of p < 0.050. RESULTS: A total of 117 brain metastases were detected in 47 patients (mean age, 59 years ± 18; 27 men). SR-DLR demonstrated significantly better lesion detection performance than DLR (mean figure of merit = 0.842 vs. 0.797; p = 0.042). Subjective image quality ratings favored SR-DLR for lesion and structure visibility, sharpness, noise, and overall quality in most cases. Objectively, SR-DLR yielded lower FWHM (1.2 mm vs. 1.9 mm; p < 0.001), higher ERS (791.3 mm- 1 vs. 645.3 mm- 1; p = 0.013) indicating enhanced sharpness as well as improved CNR (27.5 vs. 24.9; p < 0.001) compared to DLR. CONCLUSION: Compared to DLR, SR-DLR significantly enhances brain MRI quality and improves detection of metastatic lesions.
4 Apr 2026
Read appraisal →Ultrasound in obstetrics & gynecology : the official journal of the International Society of Ultrasound in Obstetrics and Gynecology
Fetal cardiac remodeling in second trimester in pregnancies with pre-eclampsia and/or fetal growth restriction: deep-learning-based approach using population-wide data
OBJECTIVE: Pre-eclampsia (PE) and fetal growth restriction (FGR) have been shown to impact fetal cardiac remodeling in the third trimester and postnatally, with evidence to suggest that this leads to increased long-term cardiovascular risk. This study aimed to examine fetal cardiac remodeling in pregnancies with PE and/or FGR during the second trimester. METHODS: This was a retrospective nationwide cohort study of ultrasound images from routine screening at 18-22 weeks of gestation in Denmark. We included 139 430 singleton pregnancies that were conceived spontaneously for which at least one relevant standardized cardiac plane (four-chamber view and/or three-vessel view) was available from the second-trimester scan. An artificial intelligence model performed automated segmentation and extraction of fetal cardiac biometric measurements. Data were collected for four groups: the reference group, defined as appropriate-for-gestational-age (AGA) fetuses from a normotensive pregnancy (n = 131 150); AGA fetuses from a PE pregnancy (AGA + PE group; n = 3095); FGR fetuses from a normotensive pregnancy (normotensive + FGR group; n = 4572); and FGR fetuses from a PE pregnancy (PE + FGR group; n = 613). Differences between groups in cardiac biometric measurements were presented as Z-scores adjusted for abdominal circumference at the second-trimester scan. RESULTS: Compared with the reference group, fetuses in the PE + AGA group had significantly larger right-sided heart structures, and the main pulmonary artery (MPA) and ventricular septum were wider. Fetuses in the normotensive + FGR group displayed significantly smaller cardiac structures, including smaller MPA and ascending aorta diameters, and increased left ventricular sphericity compared with the reference group. Fetuses in the PE + FGR group differed significantly from the reference group only in terms of smaller left atrial area, narrower ascending aorta and increased left ventricular sphericity. However, PE + FGR fetuses had significantly wider MPA and ventricular septum and larger right-sided structures compared with normotensive + FGR fetuses, mimicking the features of the PE + AGA group, suggesting a distinct effect of PE on fetal cardiac remodeling. CONCLUSIONS: Cardiac remodeling is evident as early as the second trimester, with distinct patterns of modulation in fetuses from pregnancies affected by PE vs FGR. In particular, PE is associated with an enlarged right heart and greater MPA diameter. Understanding these early differences in fetal cardiac morphology may improve risk prediction models and guide early intervention to improve long-term cardiovascular outcome. © 2025 The Author(s). Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.
4 Apr 2026
Read appraisal →Computer methods and programs in biomedicine
Application of artificial intelligence in colonoscopy imaging for polyp analysis-A systematic review
The colon is a major component of the digestive system, so early detection of colorectal polyps is essential in preventing colorectal cancer, which is a leading cause of cancer-related death worldwide. While colonoscopy remains the gold standard for polyp detection, its diagnostic accuracy is highly operator-dependent. Recent advances in Deep Learning (DL), a branch of Artificial Intelligence (AI), have shown substantial potential to improve colonoscopy image analysis by enhancing the accuracy, consistency, and objectivity of polyp detection, segmentation, and classification. Artificial intelligence-based systems have significantly reduced inter-observer variability and increased diagnostic efficiency, ultimately transforming the landscape of colorectal lesion assessment. This survey provides a comprehensive and critical analysis of the current status of deep learning applications in colorectal polyp analysis. We systematically review state-of-the-art methodologies across various DL architectures-including Convolutional Neural Networks (CNNs), transformer-based models, and hybrid approaches-and examine their performance on publicly available benchmark datasets. Additionally, we highlight the strengths and limitations of existing techniques, explore the clinical relevance of AI-assisted tools, and identify prevailing challenges such as data imbalance, real-time deployment, and generalizability across diverse populations and colonoscopy devices. By consolidating key advances and outlining future research directions, this review aims to serve as a valuable resource for researchers, clinicians, and developers seeking to leverage deep learning to enhance colorectal polyp detection, diagnosis, and clinical decision-making.
4 Apr 2026
Read appraisal →Pediatric radiology
Data mining in pediatric radiology in the era of artificial intelligence
Data mining is the systematic process of extracting useful knowledge from large multimodal datasets and is increasingly enabled by artificial intelligence (AI) methods. Pediatric radiology is a natural field for data mining because multimodal data sources, including images, reports, metadata, and electronic health records, together capture rich information on anatomy, disease, treatment, and outcomes. In the current era, the boundaries between data mining and AI are increasingly blurred. AI assists in key steps of the mining workflow through automated labeling, information extraction, and representation learning, while data mining provides the high-quality curated datasets that underpin model performance, generalizability, and safety. This review, therefore, examines both domains together, emphasizing their interdependence in the pediatric context. We describe core concepts and workflows of data mining in pediatric radiology, including data collection, linkage, annotation, analysis, validation, and governance, and outline how modern AI tools such as deep learning, large language models, multimodal fusion, and federated learning support advanced pattern discovery across limited and heterogeneous pediatric datasets. We summarize current and emerging clinical applications across diagnosis, prognosis, radiation dose monitoring, operational analytics, reporting safety nets, and continual learning. We then discuss current challenges related to data quality and standardization, ethics, regulation, workflow integration, resource disparities, sustainability, and explainability. Finally, we highlight future perspectives, including synthetic data generation, foundation models, structured reporting, and pediatric-focused ethical frameworks that aim to enable safe, transparent, and equitable integration of AI-driven data mining to improve outcomes in children.
4 Apr 2026
Read appraisal →Diagnostic and interventional imaging
Deep learning-based image reconstruction significantly improves image quality of MRI examinations of the orbit at 3 Tesla.
PURPOSE: The purpose of this study was to assess the benefit of a deep learning-based image reconstruction (DLBIR) for improving image quality in orbital magnetic resonance imaging (MRI) at 3 Tesla (T). MATERIALS AND METHODS: Seventy-one patients (48 women and 23 men) with a mean age of 52 ± 19.5 (standard deviation [SD]) years (age range: 7-90 years) who underwent MRI examination of the orbit at 3 T between January and June of 2024, were included in the study. Coronal T2-weighted MR images obtained in 70 patients and post-contrast fat-saturated (FS) coronal T1-weighted MR images obtained in 25 patients, were reconstructed with and without DLBIR, resulting in four imaging sets. Two radiologists independently and blindly measured the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the optic nerves on the four imaging sets. Image quality and orbital abnormalities were assessed using a standardized 5-point Likert scale. Comparisons between MR images obtained with and without DLBIR were performed using Wilcoxon test for ordinal and quantitative variables and McNemar test for paired binary data. RESULTS: SNR and CNR of coronal T2-weighted MR images were significantly greater using DLBIR (26.67 ± 9.03 [SD], and 14.87 ± 10.31 [SD], respectively) than without DLBIR (18.91 ± 7.28 [SD], and 9.78 ± 8.47 [SD], respectively) (P < 0.001). There were no differences in SNR and CNR between post-contrast FS T1-weighted images obtained with DLBIR (85.56 ± 63.13 [SD], and 64 ± 41.38 [SD], respectively) and those obtained without DLBIR (91.36 ± 48.49 [SD], and 43.25 ± 20.4 [SD], respectively) (P = 0.35, and P = 0.14, respectively). Qualitatively, good-to-excellent image quality was obtained more frequently with DLBIR than without DLBIR for T2-weighted and post-contrast FS T1-weighted images with respect to optic nerve sharpness (67 % vs. 16 %, and 8 % vs. 0 %, respectively), brain sharpness (90 % vs. 6 %, and 68 % vs. 4 %, respectively), and overall image quality (73 % vs. 1 % and 36 % vs. 0 %, respectively) (all P ≤ 0.001). No significant differences in the detection rates of orbital abnormalities were found between MR images obtained with and without DLBIR, including optic nerve hyperintensity (34 % vs. 31 %, respectively; P = 0.16) and optic nerve atrophy (33 % for both) on T2-weighted images, and optic nerve enhancement on post-contrast FS T1-weighted images (16 % for both). CONCLUSION: DLBIR significantly improves image quality of MRI examinations of the orbit at 3 T, without losing clinically relevant information.
4 Apr 2026
Read appraisal →European radiology
Large language model-based uncertainty-adjusted label extraction for artificial intelligence model development in upper extremity radiography
OBJECTIVES: To evaluate GPT-4o's zero-shot ability to extract structured diagnostic labels (with uncertainty) from free-text radiology reports and to test how these labels affect multi-label image classification of musculoskeletal radiographs. MATERIALS AND METHODS: This retrospective study included radiography series of the clavicle (n = 1170), elbow (n = 3755), and thumb (n = 1978). After anonymization, GPT-4o filled out structured templates by indicating imaging findings as present ("true"), absent ("false"), or "uncertain." To assess the impact of label uncertainty, "uncertain" labels of the training and validation sets were automatically reassigned to "true" (inclusive) or "false" (exclusive). Label-image pairs were used for multi-label classification using the ResNet50 architecture. Label extraction accuracy was manually verified on internal (clavicle: n = 233, elbow: n = 745, thumb: n = 393) and external test sets (n = 300 for each). Performance was assessed using macro-averaged receiver operating characteristic (ROC) area under the curve (AUC), precision, recall curves, sensitivity, specificity, and accuracy. AUCs were compared with the DeLong test. RESULTS: Automatic extraction was correct in 98.6% (60,618 of 61,488) of labels in the test sets. Across anatomic regions, label-based model training yielded competitive performance measured by macro-averaged AUC values for inclusive (e.g., elbow: AUC = 0.80 (range, 0.62-0.87)) and exclusive models (elbow: AUC = 0.80 (range, 0.61-0.88)). Models generalized well on external datasets (elbow (inclusive): AUC = 0.79 (range, 0.61-0.87); elbow (exclusive): AUC = 0.79 (range, 0.63-0.89)). No significant differences were observed across labeling strategies or datasets (p ≥ 0.15). CONCLUSION: GPT-4o extracted labels from radiologic reports to train competitive multi-label classification models with high accuracy. Detected uncertainty in the radiologic reports did not influence the performance of these models. KEY POINTS: Question Can GPT-4o automatically extract high-accuracy, uncertainty-aware diagnostic labels from routine radiologic reports of the clavicle, elbow, and thumb for use in training multi-label image classifiers? Findings GPT-4o extracted labels with > 98% accuracy, and multi-label classifiers for clavicle, elbow, and thumb radiographs performed consistently regardless of how uncertainty was handled. Clinical relevance Automated GPT-4o-based labeling of routine clavicle, elbow, and thumb radiologic reports enables the rapid conversion of radiologic reports into structured multi-label training datasets, supporting scalable development of dedicated image classification models.
4 Apr 2026
Read appraisal →Journal of biomedical informatics
Natural language generation in healthcare: A review of methods and applications
OBJECTIVE: This study presents a systematic review of natural language generation (NLG) methods and applications in the medical domain, providing quantitative and qualitative analyses to answer four key research questions regarding methods, evaluation, applications, and challenges of NLG in healthcare. METHODS: We searched PubMed, ACM Digital Library, Web of Science, Science Direct, Scopus, Embase, and ACL Anthology for NLG-related studies in healthcare from 2018 to 2024. Out of 3,988 research articles, 113 met the inclusion criteria and were analyzed across data modality, model architecture, evaluation metrics, and application domain. RESULTS AND CONCLUSION: NLG in healthcare has grown substantially, with annual publications increasing from 2 in 2018 to 40 in 2024. Of the 113 included studies, text-to-text generation was the most common data modality (65.5%), followed by image-to-text (19.5%) and multimodal-to-text (15.0%). Transformer-based architectures were dominant, especially encoder-decoder models (61.6%). Automatic evaluation metrics such as ROUGE (81.4%) and BLEU (57.5%) were widely used. Human evaluation metrics, such as Likert scales (31.9%), were increasingly adopted. The four most prevalent application domains include summarization (e.g., discharge summaries, radiology reports), clinical documentation, medical dialogue, and data augmentation. The transformer-based large language models (LLMs) and the accumulation of large-scale multimodal clinical datasets have remarkably advanced NLG in healthcare. However, challenges remain in factual consistency, explainability, evaluation robustness, and AI safety. Addressing these challenges is essential for the adoption of NLG in various healthcare applications.
4 Apr 2026
Read appraisal →Physiological measurement
StructEIT: realistic 3D EIT model generation from CT scans for deep learning applications
Objective.Artificial intelligence has significantly improved image reconstruction quality across various medical imaging modalities. However, its application in electrical impedance tomography (EIT) reconstruction remains limited, mainly due to the absence of comprehensivein vivodatasets that incorporate realistic anatomical geometries and conductivity distributions. This limitation constrains the development of supervised and data-driven reconstruction methods.Approach.To address this bottleneck, we developed StructEIT, an integrated EIT modeling framework for generating anatomically and biophysically realistic EIT simulation models. The framework incorporates three key components: (1) a structure extraction module, which automatically processes human CT scans to extract body contours and organ boundaries, thereby providing high-fidelity spatial geometry for three-dimensional finite element modeling; (2) a surface electrode attachment module, which enables flexible and accurate placement of electrodes on irregular body surfaces, supporting diverse configurations and ensuring precise definition of the electrode-tissue interface; and (3) a tissue property assignment module, which establishes frequency-dependent conductivity models for multiple organs, enabling physiologically realistic conductivity values across tissues.Main results and Significance.By bridging the gap between CT imaging and EIT, StructEIT facilitates flexible, realistic, and scalable generation of high-resolution EIT datasets. Using this framework, we constructed Chest-EIT, a thoracic EIT simulation dataset comprising over 1400 publicly available CT cases, with multiple electrode configurations provided for each case.
4 Apr 2026
Read appraisal →IEEE transactions on medical imaging
EviVLM: When Evidential Learning Meets Vision Language Model for Medical Image Segmentation
The disparity between image and text representations, often referred to as the modality gap, remains a significant obstacle for Vision Language Models (VLMs) in medical image segmentation. This gap complicates multi-modal fusion, thereby restricting segmentation performance. To address this challenge, we propose Evidence-driven Vision Language Model (EviVLM)-a novel paradigm that integrates Evidential Learning (EL) into VLMs to systematically measure and mitigate the modality gap for enhanced multi-modal fusion. To drive this paradigm, an Evidence Affinity Map Generator (EAMG) is proposed to collect complementary cross-modal evidences by learning a global cross-modal affinity map, thus refining modality-specific evidence embedding. An Evidence Differential Similarity Learning (EDSL) is further proposed to collect consistent cross-modal evidences by performing Bias-Variance Decomposition on differential matrix derived from bidirectional similarity matrices between image and text evidence embeddings. Finally, the subjective logic is used for mapping the collected evidences to opinions, and the Dempster-Shafer's theory based combination rule is introduced for opinion aggregation, thereby quantifying the modality gap and facilitating effective multi-modal integration. Experimental results on three public medical image segmentation datasets validate that the proposed EviVLM can achieve state-of-the-art performance. Code is available at: https://github.com/QingtaoPan/EviVLM.
4 Apr 2026
Read appraisal →European radiology
Deep learning-based segmentation of T1 and T2 cardiac MRI maps for automated disease detection.
OBJECTIVES: Parametric tissue mapping enables quantitative cardiac tissue characterization but is limited by inter-observer variability during manual delineation. Traditional approaches relying on average relaxation values and single cutoffs may oversimplify myocardial complexity. This study evaluates whether deep learning (DL) can achieve segmentation accuracy comparable to inter-observer agreement, explores the utility of statistical features beyond mean T1/T2 values, and assesses whether machine learning (ML) combining multiple intensity-based statistical features enhances disease detection. MATERIALS AND METHODS: T1 and T2 maps were manually segmented. The test subset was independently annotated by two observers, and inter-observer variability was assessed. A DL model was trained to segment the left-ventricular blood pool and myocardium. Average (A), lower quartile (LQ), median (M), and upper quartile (UQ) were computed for the myocardial pixels and employed in classification by applying cutoffs or in ML. Dice similarity coefficient (DICE) and mean absolute percentage error evaluated segmentation performance. Bland-Altman plots assessed inter-user and model-observer agreement. Receiver operating characteristic analysis determined optimal cutoffs. Pearson correlation compared features from model and manual segmentations. F1-score, precision, and recall evaluated classification performance. Wilcoxon test assessed differences between classification methods, with p < 0.05 considered statistically significant. RESULTS: 144 subjects (mean age 42.2 years ± 16.1, 76 men) were split into training (100), validation (15) and evaluation (29) subsets. Segmentation model achieved a DICE of 85.4%, surpassing inter-observer agreement. Random forest applied to all features increased F1-score (92.7%, p < 0.001). CONCLUSION: DL facilitates segmentation of T1/ T2 maps. Combining multiple features with ML improves disease detection. KEY POINTS: Question Manual segmentation of myocardial T1/T2 maps is time-consuming and affected by inter-observer variability; relying on single cutoff values for diagnosis may oversimplify myocardial complexity. Findings Deep learning achieves segmentation accuracy within inter-observer agreement, while machine learning improves disease detection compared to singular cutoffs. Clinical relevance Automated segmentation and feature extraction from T1/T2 maps can enhance workflow efficiency, reduce inter-observer variability, and improve diagnostic consistency. The high recall of the machine learning model minimizes missed diagnoses, ensuring more reliable disease detection.
4 Apr 2026
Read appraisal →Abdominal radiology (New York)
Artificial intelligence in gastric cancer: a systematic review of machine learning and deep learning applications
BACKGROUND: Gastric cancer (GC) remains a major global health concern, ranking as the fifth most prevalent malignancy and the fourth leading cause of cancer-related mortality worldwide. Although early detection can increase the 5-year survival rate of early gastric cancer (EGC) to over 90%, more than 80% of cases are diagnosed at advanced stages due to subtle clinical symptoms and diagnostic challenges. Artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), has shown great promise in addressing these limitations. OBJECTIVES: This systematic review aims to evaluate the performance, applications, and limitations of ML and DL models in GC management, with a focus on their use in detection, diagnosis, treatment planning, and prognosis prediction across diverse clinical imaging and data modalities. METHODS: Following the PRISMA 2020 guidelines, a comprehensive literature search was conducted in MEDLINE, Web of Science, and Scopus for studies published between 2004 and May 2025. Eligible studies applied ML or DL algorithms for diagnostic or prognostic tasks in GC using data from endoscopy, computed tomography (CT), pathology, or multi-modal sources. Two reviewers independently performed study selection, data extraction, and risk of bias assessment. RESULTS: A total of 59 studies met the inclusion criteria. DL models, particularly convolutional neural networks (CNNs), demonstrated strong performance in EGC detection, with reported sensitivities up to 95.3% and Area Under the Curve (AUCs) as high as 0.981, often exceeding expert endoscopists. CT-based radiomics and DL models achieved AUCs ranging from 0.825 to 0.972 for tumor staging and metastasis prediction. Pathology-based models reported accuracies up to 100% for EGC detection and AUCs up to 0.92 for predicting treatment response. Cross-modality approaches combining radiomics and pathomics achieved AUCs up to 0.951. Key challenges included algorithmic bias, limited dataset diversity, interpretability issues, and barriers to clinical integration. CONCLUSION: ML and DL models have demonstrated substantial potential to improve early detection, diagnostic accuracy, and individualized treatment in GC. To advance clinical adoption, future research should prioritize the development of large, diverse datasets, implement explainable AI frameworks, and conduct prospective clinical trials. These efforts will be essential for integrating AI into precision oncology and addressing the increasing global burden of gastric cancer.
4 Apr 2026
Read appraisal →Current opinion in neurology
AI in epilepsy neuroimaging
PURPOSE OF REVIEW: Recent advances in the capabilities and usability of artificial intelligence (AI) architectures coupled with increased availability of neuroimaging datasets has fuelled a rapid expansion in AI applications to epilepsy neuroimaging. This review summarizes the main applications of AI in epilepsy neuroimaging and suggests future directions for the field. RECENT FINDINGS: A range of different machine learning approaches, from multi-layer perceptrons to volumetric and graph-based convolutional neural networks, have been utilized for prediction of whether people will have epilepsy, detection of structural epilepsy lesions, localization of seizure onset zones, segmentation of resection cavities after epilepsy surgery as well as for image enhancement. SUMMARY: AI in epilepsy neuroimaging research has primarily focussed on lesion detection and localization, with a number of open and validated tools now available for evaluation across diverse settings. Additional applications of AI in epilepsy neuroimaging are either at earlier stages of development or emerging as new challenges. As these tools and their supporting evidence mature, further work addressing the hurdles of clinical integration is required.
4 Apr 2026
Read appraisal →European radiology
Deep learning models for radiography body-part classification and chest radiograph projection/orientation classification: a multi-institutional study.
OBJECTIVES: Large-scale radiographic datasets often include errors in labels such as body parts or projection, which can undermine automated image analysis. Therefore, we aimed to develop and externally validate two deep-learning models-one for categorising radiographs by body part, and another for identifying projection and rotation of chest radiographs-using large, diverse datasets. MATERIALS AND METHODS: We retrospectively collected radiographs from multiple institutions and public repositories. For the first model (Xp-Bodypart-Checker), we included seven categories (Head, Neck, Chest, Incomplete Chest, Abdomen, Pelvis, Extremities). For the second model (CXp-Projection-Rotation-Checker), we classified chest radiographs by projection (anterior-posterior, posterior-anterior, lateral) and rotation (upright, inverted, left rotation, right rotation). Both models were trained, tuned, and internally tested on separate data, then externally tested on radiographs from different institutions. Model performance was assessed using overall accuracy (micro, macro, and weighted) as well as one-vs.-all area under the receiver operating characteristic curve (AUC). RESULTS: In the Xp-Bodypart-Checker development phase, we included 429,341 radiographs obtained from Institutions A, B, and MURA. In the CXp-Projection-Rotation-Checker development phase, we included 463,728 chest radiographs from CheXpert, PadChest, and Institution A. The Xp-Bodypart-Checker achieved AUC values of 1.00 (99% CI: 1.00-1.00) for all classes other than Incomplete Chest, which had an AUC value of 0.99 (99% CI: 0.98-1.00). The CXp-Projection-Rotation-Checker demonstrated AUC values of 1.00 (99% CI: 1.00-1.00) across all projection and rotation classifications. CONCLUSION: These models help automatically verify image labels in large radiographic databases, improving quality control across multiple institutions. KEY POINTS: Question This study examines how deep learning can accurately classify radiograph body parts and detect chest projection/orientation in large, multi-institutional datasets, enhancing metadata consistency for clinical and research workflows. Findings Xp-Bodypart-Checker classified radiographs into seven categories with AUC values of over 0.99 for all classes, while CXp-Projection-Rotation-Checker achieved AUC values of 1.00 across all projection and rotation classifications. Clinical relevance Trained on over 860,000 multi-institutional radiographs, our two deep-learning models classify radiograph body-part and chest radiograph projection/rotation, identifying mislabeled data and enhancing data integrity, thereby improving reliability for both clinical use and deep-learning research.
4 Apr 2026
Read appraisal →Ann Anat
Epiphyseal interfaces.
This work explores interconnected articular and growth plate osteochondral interfaces, via complementary anatomical, biomechanical and structural analyses that combine benchtop micro-computed tomography (microCT) and high-resolution synchrotron CT (sCT) imaging approaches. We show that bony sclerosis, well-documented in late-stage OA in humans, and the osteoarthritic STR/Ort mouse model, is preceded by the amplified regionalised expansion of articular calcified cartilage (ACC) where vulnerability load-induced failure is greatest. We further observed marked alterations in the number and size of resident osteocyte lacunae in pre-OA and OA regions in comparison to controls, which age healthily, underscoring the value of CT-based approaches for elucidating early OA-related changes. Notably, our findings imply that excessive ACC formation is an early consequence of disrupted endochondral ossification that precedes and potentially drives OA sclerosis, osteophytosis and indeed, articular cartilage lesions. By applying digital volume correlation (DVC) to sCT images of loaded murine knee joints under physiological conditions, we demonstrate that the joints from a healthy parental control strain of CBA mice efficiently dissipate compressive strain, whereas OA-prone STR/Ort exhibit high strain foci in the epiphyseal compartment prior to cartilage degeneration; suggesting that early strain accumulation may therefore promote early joint vulnerability and subsequently prime degenerative progression with age. Finally, we examined how the cartilaginous growth plate (GP), responsible for all longitudinal bone growth, contributes to mechanical epiphyseal stability for articulation. Analysis of mineralised trans-physeal GP structures (bridges), revealed their dynamic and sexually dimorphic character as well as their regulation by mechanical loading and osteotropic agents. DVC analyses of sCT-imaged murine knee joints subjected to identical physiologically oriented load, identified that regionally heterogeneous load-induced strain patterns in the GP are spatially synchronised to bridge location; a relationship substantiated by finite element simulation. Together, these findings establish intra-physeal GP bridges as regulatable and mechanically sensitive structures that coordinate the transfer of local GP strains, while enhancing epiphyseal stability by distributing compressive and shear strains to thereby permit cessation of longitudinal growth.
1 Apr 2026
Read appraisal →Curr Opin Pulm Med
Foreign body removal
PURPOSE OF REVIEW: Airway foreign body aspiration remains a potentially life-threatening emergency, predominantly affecting children under 5 years and adults over 65 years. This review synthesizes current evidence on diagnostic strategies, bronchoscopic extraction techniques, procedural outcomes, complication management, and emerging technologies in airway foreign body management. RECENT FINDINGS: Multidetector computed tomography with three-dimensional reconstruction has significantly improved diagnostic accuracy, achieving sensitivity of 98-99% for radiopaque objects and 85-92% for radiolucent materials. Flexible bronchoscopy has evolved from a diagnostic tool to a first-line therapeutic modality, with recent pediatric meta-analyses demonstrating 87% success rates and adult series showing comparable outcomes to rigid bronchoscopy for appropriately selected cases. Rigid bronchoscopy maintains superiority in asphyxiating presentations, and for large (>1.5 cm), sharp, or severely impacted foreign bodies. Technological innovations including robotic-assisted bronchoscopy, electromagnetic navigation systems, and artificial intelligence-powered imaging analysis are enhancing procedural precision and safety. SUMMARY: Successful airway foreign body management requires individualized, multidisciplinary approaches integrating patient clinical status, foreign body characteristics, and institutional expertise. Success depends on appropriate bronchoscopic modality selection, comprehensive preprocedural planning, availability of specialized retrieval instruments, and readiness to manage potential complications. Integration of advanced imaging, simulation-based training protocols, and telemedicine consultation are becoming essential components of contemporary practice.
1 Jan 2026
Read appraisal →