Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 18 appraisals
PloS one
Mumtathil: Automatic PDPL compliance identification system of Arabic privacy policies documents.
In the digital era, the extensive collection and use of personal data by websites and applications has prompted various countries to implement data protection laws. The Kingdom of Saudi Arabia introduced the Personal Data Protection Law (PDPL) to regulate the handling of personal data. However, manually assessing compliance with these laws is time-consuming, and costly. This paper presents an automated system named Mumtathil (مُمتثِل) that evaluates the compliance of Arabic privacy policies with the PDPL using multi-label classification (MLC) techniques. Our experiments demonstrated that the Support Vector Machine (SVM) outperformed other machine learning models and transformer-based models like CamelBERT and AraBERT, achieving an F-score of 92%. The system addresses dataset imbalance by augmenting the Saudi Privacy Policy Dataset with recent privacy policies and synthetic data generated using ChatGPT. The Mumtathil tool simplifies compliance assessment, reducing time, cost, and effort while enhancing transparency and data protection in the digital landscape.
12 June 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 →JMIR mental health
When AI Colludes: Clinical Reliability of Training and Preference Data as a Trustworthy-AI Criterion
Research on artificial intelligence (AI) and mental health has focused largely on harms at deployment, including chatbot safety, sycophancy, and AI-associated delusions. Less attention has been paid to a prior question: whether the human-generated text and preference judgments that shape large language models are themselves clinically reliable, particularly when self-report may be distorted. This Viewpoint aims to develop the clinical psychiatric construct of collusion-the uncritical acceptance of an unreliable account-as an analytic lens for AI training and deployment, and to argue that the clinical reliability of training and preference data should be treated as an explicit trustworthy-AI criterion in mental-health-relevant systems. A conceptual synthesis of psychiatry, clinical psychology, and AI safety literature was undertaken. The analysis distinguishes three pipeline layers: pretraining corpora, preference data and posttraining methods, and deployment-time interaction. It maps the clinical construct of collusion against adjacent technical concepts, including sycophancy, reward overoptimization, grounding, refusal training, red-teaming, and live monitoring. The synthesis suggests that collusion-like dynamics are least applicable at the pretraining layer and most applicable at the preference-data and deployment layers, where unassessed user or labeler input can be reinforced without corroboration. Existing mitigations, including data curation, Constitutional AI, reward-model evaluation, grounded generation, refusal training, red-teaming, and postdeployment monitoring, address parts of this problem. However, these approaches are not yet organized around a clinically informed account of when self-report is unreliable. The central novelty is therefore not a generic claim about bias, but the proposal that clinical self-report reliability should be assessed as a distinct data-quality and governance dimension. Trustworthy-AI frameworks for mental-health-relevant applications should incorporate clinical expertise in self-report reliability into preference-data design, red-teaming, and postmarket surveillance. Adding the clinical reliability of training and preference data as an explicit criterion could complement existing technical safeguards while leaving empirical evaluation of clinician involvement as an open research agenda.
28 May 2026
Read appraisal →Acta oncologica (Stockholm, Sweden)
KaSARi: a national framework for standardized, automated, and predictive radiotherapy in Finland.
BACKGROUND AND PURPOSE: National radiotherapy (RT) data infrastructures are emerging to support treatment quality assurance and outcome research. However, prospective nationwide integration of full DICOM-RT data with toxicity and patient-reported outcome measures (PROMs) remains uncommon. KaSARi (Key advances in Standardizing, Automating, and predicting Risks in RT) was established to create national research infrastructure for prospective RT data collection and integration in Finland. MATERIALS AND METHODS: KaSARi is a prospective, consent-based multicenter cohort infrastructure involving Finnish university and central hospitals. All adult patients receiving RT are eligible. DICOM-RT datasets from treatment planning and oncology information systems are linked to clinician-graded toxicity, PROMs and relevant clinical variables. The infrastructure builds on previously validated multi-institutional DICOM workflows and follows FAIR (Findable, Accessible, Interoperable, Reusable) data principles. Predefined work packages support data harmonization, scalable data processing and enable downstream applications such as AI-based segmentation, dose prediction, and outcome modeling. RESULTS: Patient recruitment started in May 2025. By February 2026, 221 patients had been enrolled at the coordinating center, initially including breast cancer patients and subsequently expanding to all RT indications. Recruitment at a second center began in November 2025, with 37 patients enrolled. The infrastructure is projected to include approximately 29,000 patients over 15 years. Integration of dosimetric, clinical and PROM data is progressing through a phased national implementation as additional centers join the network. INTERPRETATION: KaSARi represents a national infrastructure prospectively integrating full DICOM-RT datasets with clinician-graded toxicity and PROMs across all RT indications within a unified research protocol. The study provides a foundation for nationwide outcome modeling, harmonization of RT practices and development and validation of data-driven AI-based methods.
19 May 2026
Read appraisal →European journal of dental education : official journal of the Association for Dental Education in Europe
Assessment of the Large Language Models in Creating Dental Board-Style Questions: A Prospective Cross-Sectional Study
INTRODUCTION: Although some studies have investigated the application of large language models (LLMs) in generating dental-related multiple-choice questions (MCQs), they have primarily focused on ChatGPT and Gemini. This study aims to evaluate and compare the performance of five of the LLMs in generating dental board-style questions. MATERIALS AND METHODS: This prospective cross-sectional study evaluated five of the advanced LLMs as of August 2024, including ChatGPT-4o (OpenAI), Claude 3.5 Sonnet (Anthropic), Copilot Pro (Microsoft), Gemini 1.5 Pro (Google) and Mistral Large 2 (Mistral AI). The five most recent clinical guidelines published by The Journal of the American Dental Association were used to generate a total of 350 questions (70 questions per LLM). Each question was independently evaluated by two investigators based on five criteria: clarity, relevance, suitability, distractor and rationale, using a 10-point Likert scale. RESULT: Inter-rater reliability was substantial (kappa score: 0.7-0.8). Median scores for clarity, relevance and rationale were above 9 across all five LLMs. Suitability and distractor had median scores ranging from 8 to 9. Within each LLM, clarity and relevance scored higher than other criteria (p < 0.05). No significant difference was observed between models regarding clarity, relevance and suitability (p > 0.05). Claude 3.5 Sonnet outperformed other models in providing rationales for answers (p < 0.01). CONCLUSION: LLMs demonstrate strong capabilities in generating high-quality, clinically relevant dental board-style questions. Among them, Claude 3.5 Sonnet exhibited the best performance in providing rationales for answers.
6 May 2026
Read appraisal →Biomedical physics & engineering express
Kali MC: an open-source toolkit for intraoperative electron radiation therapy treatment planning
Intraoperative electron radiation therapy (IOERT) is commonly delivered without intraoperative imaging, limiting patient-specific planning and requiring fast, reliable treatment calculations in the operating room. As a result, monitor unit (MU) calculations are often performed using spreadsheets and static look-up tables derived from measurements in water. We present Kali MC, an open source, Python-based software developed to streamline IOERT absorbed dose visualization, MU calculation and report generation for treatments delivered with the Liac HWL mobile accelerator. The software provides a graphical interface to visualize precomputed Monte Carlo absorbed dose distributions in water and performs MU calculations using experimentally determined output factors. Optional atmospheric pressure correction and predefined rescaling factors are supported to improve consistency of dose delivery and target coverage with the prescription isodose. Kali MC also generates customizable PDF treatment reports and exports DICOM RT Plan objects to facilitate integration with record-and-verify systems. By integrating validated dosimetric data, correction methods, and 3D dose visualization into a transparent platform, Kali MC supports efficient intraoperative decision-making; however, its use should be locally commissioned and evaluated according to institution-specific procedures and equipment characteristics.
28 Apr 2026
Read appraisal →Food chemistry
A comprehensive review on wine authentication and traceability: Current issues, solutions and future perspectives
Wine, as a commercially significant and globally consumed beverage, faces persistent challenges regarding authenticity verification and origin traceability, which threaten both consumer interests and market fairness. This paper comprehensively summarizes prevalent types of wine fraud issues and reviews modern analytical techniques including spectroscopy, chromatography, mass spectrometry, and emerging intelligent sensing techniques alongside advanced machine learning approaches for enhancing authentication and traceability of wine. The study highlights recent applications of these methods in verifying wine authenticity and geographical origins, emphasizing their evolving integration. Despite considerable progress, future development should focus on rapid detection, integrating deep learning with traditional analytical techniques, building wine fingerprint database, and leveraging blockchain for supply chain transparency. These measures are crucial to ensure the authenticity and geographical origins of wine, thereby safeguarding the interests of all stakeholders, including consumers and producers.
17 Apr 2026
Read appraisal →Journal of medical Internet research
Beyond Missingness: Systematizing Methods for Comprehensive Data Fitness Assessment in Clinical Research
Secondary use of clinical data offers unprecedented opportunities to rapidly conduct large-scale research and improve patient care. However, incomplete understanding of data quality requirements for a study often causes significant delays in executing analyses and validating results. Current practice has largely followed 2 paths. First, multi-institutional networks have developed general data quality programs, but these are typically tied to unique network characteristics and do not address study-specific requirements well. Second, models have been proposed to formalize the requirements for data fitness analyses without extending to the methods needed to meet these requirements. More recently, tools have been developed to conduct cohort-centric screening, focusing on generally applicable structural checks such as missingness or facial implausibility. These provide a first level of information but incompletely capture the fitness requirements of an analysis. In turn, investigators conduct per-study exploratory analyses, but these efforts are typically ad hoc and partially reported, which can hinder reproducible science and delay advances in patient care. Analogously to advances over the past decade in data modeling and reproducible analytics, there is a need for a more systematic, capable approach to study-specific data quality assessment (SSDQA). We discuss such a model, which guides improved SSDQA design and implementation, including metadata for consistent annotation and reporting of data quality assessment results. The model integrates theoretical principles of data quality testing with pragmatic considerations of application to clinical data, providing a consistent approach to specifying data quality assessment checks. Additionally, it proposes to regularize check application through a standard set of options. The SSDQA model builds on current practice, providing a path toward more complete, sound, and reproducible assessments. These characteristics foster multidisciplinary collaboration to identify data quality issues that, in turn, inform decisions about study design and provide important context that has a bearing on adoption of results.
17 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 →PloS one
Unsupervised industrial image defect detection based on autoencoder and GANs
In the process of intelligentization in modern manufacturing, especially in industrial fields such as automobile manufacturing, semiconductor production, and electronic product assembly, product quality control is crucial. Traditional defect detection methods face problems such as supervised learning methods relying on a large amount of labeled data, weak generalization ability, and high cost. In the process of intelligent manufacturing, industrial product quality control is a key link to ensure production safety and product consistency. Especially in typical industrial scenarios such as automobile manufacturing, semiconductor production, and electronic product assembly, traditional defect detection methods are difficult to meet the needs of actual production lines for high-precision and high-efficiency detection due to their reliance on a large amount of labeled data, weak generalization ability, and high cost. To solve the problem of defect detection in small and zero sample scenarios in these industries, improve detection sensitivity and localization accuracy, and enhance the model's generalization ability to unknown defects, a unsupervised industrial image defect detection method based on autoencoder and Generative Adversarial Networks (GANs) is proposed. This study constructs an Multi-level Deep feature Adaptive fusion AutoEncoder (MDAAE) module, extracts multi-scale features through Pre-Trained Convolutional Neural Backbone Networks (PTCNBN), introduces Attention Mechanism (AM) to dynamically calculate feature weights, and achieves feature fusion and reconstruction. Meanwhile, the self-AM is fused to improve the GANs. The self-attention module efficiently captures long-range dependencies and generates an adversarial network training objective function to optimize the generator's unsaturated loss. The results showed that the Area Under the Curve (AUC) of the Recall-Precision curve of the research method reached 93.6 ± 0.5%, and its F1-Score value exceeded 0.890 ± 0.003 in defect types such as scratches and dents. In practical applications, the inference time of the research method remained stable at 3.0 ± 0.2 GB of central processor memory, and the Fréchet Inception Distance (FID) value fully converged at the 2230th iteration, with a stable FID value of 3. The false alarm rate was only 8.6 ± 0.7% under strong light conditions. The proposed UIIDD method based on autoencoder and GANs had good robustness, generalization ability, reliability, and efficiency. This study effectively solves the problems of poor robustness, weak generalization ability, and high cost of traditional defect detection methods.
12 Apr 2026
Read appraisal →PDA journal of pharmaceutical science and technology
The European Qualified Person: Legal Accountability, Quality Leadership, and AI Integration in Modern Pharmaceutical Manufacturing: Navigating Complex Therapies and AI-Supported Oversight to Safeguard Patient Safety
The European Qualified Person (QP) holds unique legal responsibility for certifying batch release, ensuring patient safety, and maintaining regulatory compliance. Modern pharmaceutical manufacturing-especially in biologics, advanced therapy medicinal products (ATMPs), and personalized medicines-generates complex, multi-site, data-rich environments that challenge traditional oversight. Artificial Intelligence (AI) offers predictive analytics, anomaly detection, and trend recognition to support decision-making but cannot currently replace the QP's scientific judgment or legal accountability. The forthcoming Annex 22 introduces Human-in-the-Loop (HITL) frameworks that embed human oversight within AI-supported processes, aligning technological advancement with regulatory control. This review explores evolving QP responsibilities, core and emerging competencies, AI integration within Quality Risk Management (QRM) and Good Manufacturing Practice (GMP) frameworks, and ethical considerations, illustrated with practical case studies. By developing AI literacy and applying HITL oversight, QPs can more effectively translate technological potential into transparent, science-based, and patient-centered batch-release decisions.
12 Apr 2026
Read appraisal →PDA journal of pharmaceutical science and technology
Integrating Real-Time Viable Biofluorescent Particle Counters within Robotic Gloveless Isolators
Advanced technologies in both aseptic filling and environmental monitoring are coming together to improve the resilience and sterility assurance of aseptic processing. Continuous, real-time environmental monitoring using biofluorescent particle counters (BFPCs) to detect viable and nonviable particles during the aseptic filling process for injectable drugs is gaining traction as an accepted alternative to conventional methods such as active air sampling instruments. Evolving regulatory guidance, including EU Annex 1 guidelines, are increasingly recommending the adoption of isolator technology as well as continuous environmental monitoring during GMP processes, including technologies that offer real-time feedback during drug product manufacturing. Computational flow dynamics and airflow visualization studies are additional tools that support the design of equipment and determination of locations for environmental monitoring. The current nutrient culture media growth-based environmental monitoring, rooted in science from 150 years ago, is unable to keep pace with recent technological advances and the ability to immediately react to an out-of-control state. Here we examine the design of an isolator that eliminates human intervention and indirect product contact parts during aseptic fill finish operations and present computational fluid dynamics (CFD) studies verified by airflow visualization, along with the incorporation of BFPCs at critical areas. Also, we report the results of the interference study characterizing the baseline results for detection of total particles (nonviable plus viable) using a BFPC within a robotic gloveless isolator during dynamic and static operating conditions. The results of our study using BFPCs demonstrate that airflow in the robotic gloveless isolator provides protection of critical areas from contamination during the tub peeling process, and that the stoppering process in this environment does not generate detectable particles. During dynamic conditions and material transfer, the study demonstrates the design of the robotic gloveless isolator prevents false positives from interfering with materials during normal operations.
12 Apr 2026
Read appraisal →PDA journal of pharmaceutical science and technology
Bridging Guidance and Regulation: Interpreting the Draft Annex 22 on Artificial Intelligence in GMP Manufacturing
This paper builds upon Stassen et al. (2025), "Recommendations for Artificial Intelligence Application in Continued Process Verification," to provide a detailed interpretation of the European Medicines Agency (EMA) Draft Annex 22-Artificial Intelligence (July 2025 draft) within the context of GMP-regulated pharmaceutical manufacturing. The draft annex introduces a structured, risk-based framework for artificial intelligence (AI) adoption, defining expectations for intended use, validation, life cycle management, explainability, and human-in-the-loop (HITL) oversight. This manuscript examines the alignment between Annex 22 provisions and continued process verification (CPV) recommendations, identifies convergence, operational challenges, and areas for future development, and proposes a roadmap for compliant AI implementation. Conceptual frameworks and workflow illustrations depict the harmonization of CPV principles, life cycle management, and emerging regulatory expectations. This review supports harmonized AI governance and readiness for Annex 22 compliance across the pharmaceutical sector.
12 Apr 2026
Read appraisal →International journal of pharmaceutics
Beyond spectroscopy: Machine vision as the future of non-destructive testing in 3D-printed pharmaceuticals
3D Printing (3DP) has revolutionized personalized medicine by enabling the production of patient-specific dosage forms and adaptable drug delivery systems. However, despite these advances, 3DP presents significant challenges in quality control (QC), particularly as traditional methods may not be suitable for small-scale or on-demand manufacturing. To address these challenges, it is essential to develop flexible, precise, and efficient QC strategies that ensure product integrity. This review focuses on non-destructive spectroscopic- and imaging-based testing techniques that can bridge this gap. Techniques such as Near-Infrared Spectroscopy, Raman Spectroscopy, Terahertz Spectroscopy, Optical Coherence Tomography, and Hyperspectral Imaging provide robust alternatives to conventional QC approaches. Moreover, Machine Vision (MV) is emerging as a promising tool for real-time defect detection and visual inspection. When integrated with advanced algorithms and artificial intelligence, MV enables automated decision-making and defect identification in 3DP pharmaceuticals. Recent advancements in deep learning, especially through Convolutional Neural Networks (CNNs) and You Only Look Once (YOLO)-based models, have further enhanced the potential of MV in this domain, offering improved accuracy, speed, and adaptability under varying conditions. This review also includes a comparative evaluation of these techniques, highlighting the increasing importance of non-destructive tools, particularly MV, in supporting innovation and ensuring quality within an increasingly personalized pharmaceutical landscape. Furthermore, it also discusses existing gaps, regulatory challenges, and the need for standardization to ensure the safe and reliable integration of these technologies into pharmaceutical manufacturing workflows.
10 Apr 2026
Read appraisal →Analytica chimica acta
Point-of-need and portable sensing strategies for quality and safety assessment of Traditional Chinese Medicine products: A critical analytical review.
Sensor technology has emerged as a transformative tool for point-of-need and portable quality control and safety assessment of Traditional Chinese Medicine (TCM) products. Although separation and detection technologies have improved, there is still a strong need to critically evaluate the entire analytical workflow from sample preparation to data interpretation to meet regulatory and routine monitoring requirements. This review critically evaluates the complete analytical workflow, from sample pretreatment to intelligent data interpretation, necessary to bridge the gap between laboratory prototypes and real-world deployment. Additionally, it provides a systematic summary of the mechanisms and applications of common types of sensors, utilising nanomaterials and molecular recognition elements to overcome the specific challenge of complex TCM matrices. A particular focus is placed on the emerging roles of Artificial Intelligence (AI) and Machine Learning (ML) in resolving high-dimensional spectral data and enhancing pattern recognition accuracy. Furthermore, we critically analyze current bottlenecks, including matrix interference, the lack of standardized validation protocols, and the reliability of portable devices in uncontrolled environments. Future directions are highlighted towards the development of self-validating, cloud-integrated (IoT), and eco-friendly sensing systems, aiming to support the global standardization and safety assurance of TCM products.
9 Apr 2026
Read appraisal →Physics in medicine and biology
End-to-end testing of an online adaptive radiotherapy workflow for liver tumors using a dynamic Thoraco-abdominal anthropomorphic phantom.
Objective.Adaptive radiotherapy (ART) requires robust end-to-end (E2E) testing tools capable of reproducing organ deformation, physiological motion, and multimodal imaging properties. TheToraco-abdominalAnthropomorphic Phantom withMotion forAdaptiveRadiotherapy (TAM-ARa) is a dynamic anthropomorphic phantom developed to support the design, validation, and quality assurance of online ART workflows.Approach.The phantom was constructed with anatomically realistic bone, lung, and abdominal organ models fabricated from tissue equivalent materials. Modular components allowed simulation of interfractional and intrafractional anatomical variations, including ventilator driven respiratory motion, abdominal deformation, and variable gastric filling. CT, CBCT, and 3T MRI scans were acquired to assess imaging performance. Two E2E tests of an online adaptive intensity-modulated radiotherapy (IMRT) workflow for liver tumors were performed on a Varian Ethos (ETHOS, Varian, USA) system. Multiple deformation scenarios were investigated, including a static reference configuration, two abdominal and one thoracic deformation for ionization chamber (IC) dosimetry, and a static plus abdominal deformation scenarios for radiochromic film dosimetry.Main results.The TAM-ARa phantom demonstrated realistic radiological characteristics, with CT Hounsfield units and MRI relaxation times closely matching reportedin vivovalues. The modular architecture allowed reproducible and controlled abdominal deformations, while setup and dosimeter placement were completed within minutes. In E2E tests for adaptive IMRT of liver tumors, excellent agreement was achieved between measured and planned dose distributions after the treatment plan was adapted. IC E2E measurements showed that online adaptation consistently restored accurate dose delivery under deformation, reducing deviations to below 3%. Independent film measurements confirmed sub-millimeter geometric accuracy and full planning target volume coverage after adaptation, demonstrating effective compensation of deformation-induced errors.Significance.The TAM-ARa phantom provides a versatile and reproducible platform for multimodal E2E testing and validation of online ART workflows. Employing its realistic anatomical design, modular structure, and motion capabilities, TAM-ARa was successfully applied for E2E tests to validate a workflow for adaptive IMRT of liver tumors.
8 Apr 2026
Read appraisal →Clinical chemistry
Discovery of an Artificial Intelligence Label Feedback Loop: How the Success of a Clinically Implemented Artificial Intelligence Algorithm Has Created an Unforeseen Challenge to Algorithm Retraining
BACKGROUND: Artificial intelligence (AI) augmented laboratory tests can improve quality, efficiency, cost-effectiveness, and staff satisfaction. However, the clinical success of these tests can introduce unforeseen challenges for model retraining. This study describes the discovery of an "AI label feedback loop" in a clinically implemented AI-augmented kidney stone composition test. METHODS: An AI-augmented kidney stone composition test (V1) has been previously deployed for clinical kidney stone characterization. After several years of clinical use, a retrained model (V2) was developed using 6 times more data. Model performance of both V1 and V2 were evaluated across 3 datasets: a recent production validation (hold-out) set (mostly AI-influenced labels), the original V1 validation set (pre-AI, entirely human-labeled), and a subset of recent cases with exclusively human-generated or human-corrected labels. RESULTS: V2 demonstrated a 10% lower concordance rate than V1 when evaluated on the recent production hold-out set, despite a much larger training dataset. Performance between V1 and V2 was similar when applied to the pre-AI validation set. Notably, V2 outperformed V1 on the recent subset of cases with human-only or human-corrected labels, particularly for less-common stone types. These findings revealed an AI label feedback loop, confounding retraining and evaluation. CONCLUSIONS: The integration of AI into clinical practice can potentially influence reported test results, complicating the development and evaluation of future models. To mitigate AI label feedback loops, ongoing human annotation and careful validation set construction are essential. These strategies can ensure reliable performance assessment and support the safe evolution of clinical AI systems.
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 →