Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 44 appraisals
Sleep medicine reviews
Machine and deep learning in REM sleep behavior disorder: a scoping review and analysis of reporting quality
Rapid eye movement (REM) sleep behavior disorder (RBD) is a parasomnia, and its isolated form is of particular interest, as it is an early phase alpha-synucleinopathy. Machine learning (ML) and deep learning (DL) models offer potential for automated detection, prediction of phenoconversion, and phenotyping. This scoping review identified 75 studies applying ML/DL in RBD and evaluated their methodological and reporting quality using the APPRAISE-AI tool. Most studies (73.3%) focused on RBD detection and mainly used polysomnographic data for this, while 16% addressed prediction of phenoconversion, with imaging data being the most employed modality. Sample sizes were generally small (most studies including only 20-100 individuals). According to APPRAISE-AI scores, 80% of studies had moderate overall methodological and reporting quality. Common deficiencies included lack of transparency in data and code sharing (23.3%), and poor reporting of hyperparameter tuning (17.1%), bias assessment (26.9%), and error analysis (0.66%). Data leakage was observed in 32% of studies. These issues hinder clinical translation and prevent incremental progress between research groups. Without transparent reporting and shared resources, replication and model comparisons become nearly impossible. Future work should adopt open science principles and rigorous validation to advance AI-based tools in sleep medicine.
3 Aug 2026
Read appraisal →Neurological research
Federated deep learning model for epilepsy seizure detection using electroencephalogram (EEG) signal
OBJECTIVES: Epilepsy is a chronic neurological disorder characterized by recurrent seizures due to abnormal brain activity, which affects individuals' health and quality of life. Traditional seizure detection methods face challenges related to data privacy and security as well as difficulty in fully capturing both temporal and spatial relationships within the Electroencephalography signal. To address these limitations, a Federated Learning Enabled Unified Transformer model is proposed. METHODS: The Federated Learning with Paillier Homomorphic Encryption is deployed for preserving data privacy and enabling collaborative model training. Adaptive noise filtering and Independent Component Analysis are deployed to remove the noise and artifacts. The Multi-Scale Wavelet Coefficient is applied for signal decomposition that effectively extracts seizure-related features by decomposing Electroencephalography signals into multiple sub-bands. The Hybrid Graph-Based Attention Framework integrated Edge-Enhanced Graph Convolutional Networks for spatial feature extraction and Spectral Graph Attention for frequency-based feature selection, and refining feature representation for improving the classification accuracy. Further, the proposed technique utilizes a Unified Transformer model for seizure classification that efficiently captures temporal and spatial dependencies in Electroencephalography signals. RESULTS: The proposed model is validated on three datasets and attains 98.91% accuracy, 98.93% security, and 98.82% precision. The simulation outcomes indicate that the Federated Learning Enabled Unified Transformer model achieved outstanding performance when compared to existing models. DISCUSSION: The Federated Learning Enabled Unified Transformer model provided a superior performance by integrating Federated learning and hybrid Deep Learning models. It ensures that the model is more suitable for healthcare users.
2 Aug 2026
Read appraisal →Journal of psychiatric research
Systematic review of machine learning and deep learning models for EEG-based detection of depression.
OBJECTIVE: Depression is a leading cause of global disability, motivating the development of objective and scalable diagnostic approaches. Quantitative electroencephalography (QEEG) combined with machine learning (ML) and deep learning (DL) techniques has gained increasing attention for depression detection. This systematic review aimed to critically examine and descriptively compare the methodologies, performance metrics, and limitations of ML- and DL-based models applied to EEG data for depression detection. METHODS: A systematic review was conducted in accordance with PRISMA 2020 guidelines. Seven electronic databases (PubMed, Scopus, IEEE Xplore, ScienceDirect, Web of Science, SAGE Journals, and MDPI) were searched for peer-reviewed studies published between 2020 and 2024. Eligible studies included human participants, used EEG signals for depression detection, and applied ML or DL algorithms. Extracted information comprised algorithm type, sample size, EEG acquisition parameters, validation strategies, and reported performance metrics, which were synthesized descriptively across studies. Risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool. RESULTS: A total of 42 studies met the inclusion criteria, including 23 ML-based and 19 DL-based investigations. Reported classification accuracy ranged from approximately 76% to 100%. DL studies showed a higher mean reported accuracy than ML studies (93.92% vs. 90.78%); however, this difference was not statistically significant in the exploratory non-parametric comparison. Near-perfect performance values were frequently observed in studies with small sample sizes and subject-dependent or exclusively internal validation strategies, raising concerns regarding overfitting and limited generalizability. Studies relying on publicly available datasets tended to report more stable performance. QUADAS-2 assessment revealed recurrent risk-of-bias concerns, particularly in the domains of patient selection and index test conduct. CONCLUSIONS: Both ML and DL approaches demonstrate potential for EEG-based depression detection, but reported performance differences between them should be interpreted cautiously. Although DL studies tended to report higher accuracy values, this pattern was not statistically significant in exploratory analyses and was strongly influenced by sample size, validation strategy, and methodological design. Future research should prioritize larger and more diverse samples, subject-independent or external validation strategies, and standardized reporting frameworks to enhance methodological rigor and clinical applicability.
2 Aug 2026
Read appraisal →Journal of neural engineering
A roadmap to navigate the future of neural engineering
A group of leaders in neural engineering collaborated to develop a roadmap to navigate the future of neural engineering. We covered a range of themes, including brain machine interfaces, neural modeling, artificial intelligence and machine learning, neural interfaces, neural imaging, augmented rehabilitation, and neuromaterials. For each topic we reviewed the current status, identified current and future challenges, and speculated on the emerging and necessary advances in science and technology to meet these challenges. Neural engineering will continue to yield the approaches and insights that advance the diagnosis and treatment of nervous system disorders, as well as provide new understanding of neural function.
28 July 2026
Read appraisal →Journal of neural engineering
Cortical-SSM: a deep state space model for motor imagery decoding from EEG signals
Objective.Classification of electroencephalogram (EEG) signals obtained during motor imagery (MI) has substantial application potential, including for communication assistance and rehabilitation support for patients with motor impairments. These signals remain inherently susceptible to physiological artifacts (e.g. eye blinking, swallowing), which pose persistent challenges. Although Transformer-based approaches for classifying EEG signals have been widely adopted, they often struggle to capture fine-grained dependencies within them.Approach.To overcome these limitations, we propose Cortical-SSM, a novel architecture that extends deep state space models to capture integrated dependencies of EEG signals across temporal, spatial, and frequency domains. We validated our method across two large-scale public MI EEG datasets containing more than 50 subjects.Main results.Our method outperformed baseline methods on the two benchmarks. Furthermore, visual explanations derived from our model indicate that it effectively captures neurophysiologically relevant regions of EEG signals.Significance.These results indicate that Cortical-SSM provides a robust and interpretable alternative to attention-based architectures for MI EEG decoding. By enabling physiologically grounded feature learning, our method advances the reliability of subject-independent EEG classification and supports the development of practical and clinically deployable brain-computer interface systems.
26 July 2026
Read appraisal →Nanoscale
Advances in printable flexible and stretchable thin-film electrodes: materials, interfaces, technologies and bioelectronic applications
The mechanical mismatch between rigid clinical electrodes and soft biological tissues remains a primary bottleneck restricting the stability of long-term electrophysiological monitoring. Printable flexible thin-film electrodes offer a compelling solution by enabling additive, high-throughput patterning on flexible and stretchable substrates, thereby circumventing the reliance on vacuum environments and high-temperature processing typical of conventional microfabrication. This review synthesizes recent advances in functional inks, ranging from metals and carbon nanomaterials to conductive polymers and ionogels, together with high-resolution printing techniques. Addressing the critical challenge of interfacial failure in flexible devices, we explore engineering strategies to enhance adhesion at both electrode-substrate and electrode-tissue interfaces. Specifically, we analyze the pivotal roles of physical interlocking and chemical anchoring mechanisms in suppressing dynamic delamination and maintaining device integrity. Finally, the review highlights representative applications in wearable electronics, implantable systems, and emerging organoid interfaces, and outlines key translational challenges, including long-term stability and manufacturing reproducibility.
25 July 2026
Read appraisal →Journal of neural engineering
Multimodal quantification of cognitive load using a printed wearable facial bio-potential system
Objective.Cognitive load refers to the amount of mental effort required to process information and perform tasks. It has a strong impact on both learning and task performance and, therefore, plays a central role in cognitive neuroscience, psychology, neurology, human-machine interaction, and education research. Despite its significance, measuring cognitive load is challenging due to its complex and dynamic nature, influenced by task complexity, individual differences, and emotional states. While standard methods are based on subjective self-reports, numerous objective methods have been explored to assess cognitive load, such as electroencephalography (EEG), heart rate variability, and performance-based metrics. Despite these efforts, it remains challenging to objectively measure cognitive load in stationary and even more so in freely moving individuals.Approach.In this study, a wireless wearable system was employed to collect facial bio-potential markers and systematically compare their sensitivity to cognitive load. A dedicated task paradigm was developed and validated to induce graded cognitive load, allowing precise synchronization of physiological signals with task events. Facial muscle activity, eye movements, and brain activity were acquired simultaneously using a single facial electrode array.Main results.Across nine participants, entropy-based features tracked cognitive load more closely than conventional band-power measures. Facial EMG was the most informative modality: a cross-subject (leave-one-subject-out) model predicted the NASA-TLX weighted workload score with a mean Spearman correlation= 0.792 (95% CI [0.747, 0.836]; mean absolute error4.3, on a NASA-TLX 0-100 scale), and a single forehead-EMG entropy feature (the EMG-band entropy of forehead channel Ch14, above the right eyebrow) was selected in every cross-subject fold, generalizing across all participants without subject-specific calibration. An EEG-only model carried weaker but genuine cross-subject information (= 0.739), whereas EOG features did not pass the feature-selection threshold in this seated paradigm.Significance.Facial bio-potentials offer a practical, wearable system for objective cognitive load assessment. An outdoor pilot experiment further demonstrated electrophysiological measures under non-laboratory conditions. These findings offer insights into the suitability of different physiological indicators for assessing cognitive load, particularly the informativeness of facial EMG for seated assessment and the robustness of blink-based measures under ambulatory motion, contributing to the future development of practical measurement techniques in dynamic everyday settings.
23 July 2026
Read appraisal →Biomedical physics & engineering express
MELF: a multi-view ensemble learning framework for normative resting state EEG signal quality assessment
Normative resting-state electroencephalography (rsEEG), defined as standardized recordings from healthy populations, provides essential baselines for studying brain mechanisms and psychiatric disorders. However, its reliability critically depends on rigorous quality assessment, as artifacts or over-processing can distort subsequent analyses. Current approaches often rely on single-feature perspectives, which fail to capture the multidimensional nature of signal quality. We propose MELF, a quality assessment framework for normative rsEEG that systematically integrates three complementary feature domains: time domain (mobility, complexity), frequency domain (-band relative power, parallel log spectra index (PaLOSi)), and spatial domain (-band directed phase lag index, bad channel ratio, ICLabel-classified brain component proportion). A random forest-based ensemble classifier is constructed using dynamic weight allocation based on cross-validation accuracy (Acc). Experiments on four open rsEEG datasets (DSTR, TRCS, LEMON, and HBN) and expert-annotated clinical data demonstrate that: (1) MELF achieves up to 10% Acc improvement over single-view methods; (2) MELF outperforms traditional machine learning and deep learning baselines while reducing computational time, reaching 97.69% Acc (AUC = 0.9957), and maintains strong performance on an independent test set (92% Acc, AUC = 0.88); (3) the spectral view contributes the most, confirming the evaluative capability of the PaLOSi. MELF provides an interpretable, high-Acc, and robust automated framework for normative rsEEG quality assessment, with broad applications in neuroscience research and clinical quality control.
22 July 2026
Read appraisal →Journal of neural engineering
Dynamic functional graph-Laplacian priors integrated with optimization for EEG source localization
Objective. Electroencephalography (EEG) source localization is an ill-posed inverse problem in which conventional methods often rely on static anatomical or smoothness assumptions and may neglect task-related dynamic functional interactions. This study aims to develop a dynamic graph-regularized EEG source localization framework that incorporates time-varying functional connectivity directly into inverse reconstruction and improves source-space motor imagery (MI) decoding.Approach. We propose DynaGraph-alternating direction method of multipliers (DG-ADMM), a source localization framework that combines linearly constrained minimum variance (LCMV) beamforming, region-of-interest-level dimensionality reduction, dynamic phase synchronization analysis, graph-Laplacian regularization, and efficient optimization. Initial source estimates are obtained using a LCMV beamformer. Region-level source signals are then extracted using principal component analysis and sliding-window phase-locking values with surrogate-based statistical testing are used to construct reliable dynamic functional graphs. The resulting graph Laplacian is mapped back to source space and embedded as a structured prior in an ADMMs optimization problem.Main results. Experiments on the MNE sample dataset showed that DG-ADMM produced spatially concentrated and physiologically plausible source patterns. On the PhysioNet MI dataset, the proposed framework achieved a binary left-versus-right MI classification accuracy of 93.52%, outperforming representative deep learning baselines. In a 320-dataset synthetic benchmark covering single-source, double-source, correlated-source, and dynamic-source conditions at two signal-to-noise ratios, DG-ADMM achieved the lowest mean center-of-mass localization error in five of eight conditions and showed its clearest advantage for dynamic sources.Significance. The results demonstrate that dynamic functional connectivity can serve as an informative graph-structured prior for EEG inverse reconstruction. DG-ADMM provides an interpretable and computationally feasible strategy for improving spatial focus, temporal consistency, and source-space decoding performance in EEG-based brain-computer interfaces.
21 July 2026
Read appraisal →Neuroinformatics
NeuroFusion: A Unified Framework for Generalized Visual Stimulus Decoding from fMRI Across Datasets and Subjects
Recent advancements in neural decoding have shown promising results in reconstructing visual experiences from brain activity. However, existing approaches focus primarily on decoding within a single dataset or subject, which limits generalization across various sources of neuroimaging. In this work, we propose a novel framework for the decoding of visual stimuli between subjects and between data sets, integrating neural recordings from multiple publicly available fMRI datasets. To address inherent intersubject and interdataset variability, we introduce a contrastive learning-based alignment strategy using image embeddings from a pre-trained IP-Adapter model. Our approach learns a shared latent space by aligning subject-specific neural representations with image features, enabling generalized decoding across both subjects and datasets. In addition, we propose a simple yet effective data augmentation method using ridge regression. This method synthesizes realistic fMRI-like signals from novel images by predicting voxel activity and injecting learned noise distributions, thus enhancing training diversity and model robustness. To the best of our knowledge, while several recent studies have explored cross-subject decoding, we extend recent cross-subject decoding efforts by training a single unified framework jointly across multiple public fMRI datasets and subjects, enabling cross-dataset transfer in addition to cross-subject generalization. We distinguish this multi-dataset unified training setting, where each dataset contributes training data, from a stricter leave-one-dataset-out transfer setting in which the target dataset is excluded from source pretraining and used only for lightweight alignment-layer adaptation. Empirically, our unified model achieves strong semantic reconstruction across datasets (e.g., up to 94.8% CLIP similarity on NSD (AUG) and 0.403 SSIM on BOLD5000 after lightweight finetuning), demonstrating robust cross-subject and cross-dataset transfer.
20 July 2026
Read appraisal →Science advances
Reduction of appearance artifacts in wearable on-skin electronics
Facial electrophysiological signals are crucial to human-machine interfaces and health care monitoring. Soft and skin-conformable electrodes enabled long-term and comfortable signal monitoring. However, the appearance of the electrodes affects the wearer's social interactions and self-identity, making daily usage difficult and leaving appearance artifacts. Here, we developed fully invisible and unperceivable on-skin electrodes free from appearance artifacts. Neither the wearer nor observers can detect the visual and tactile presence of the electrodes on the skin. The unperceivable property was confirmed with sensory experiments and physical characterizations of the film on skin. Furthermore, our invisible electrodes did not affect the psychological conditions of the wearers, which confirms the feasibility of artifact-free monitoring in daily lives. Last, we demonstrated the functionality of our electrode with successful monitoring of various facial electrophysiological signals, including electrooculogram (EOG), electromyogram (EMG), and electroencephalogram (EEG). Our fully invisible electrodes provide a promising direction in developing on-skin bioelectronics, seamlessly integrating health monitoring and human-computer interaction technologies into people's daily lives.
19 July 2026
Read appraisal →Biomedical physics & engineering express
EEG-EMG spatiotemporal cross-attention fusion network for functional upper-limb movement classification
Objective. Electroencephalography (EEG) and electromyography (EMG) are widely used for decoding motor intentions, yet unimodal approaches often suffer from low robustness and limited representational information. EEG-EMG hybrid brain-computer interfaces (BCIs) can bridge cortical intention and muscular execution, but challenges remain in signal alignment, fusion modeling, and clinical generalization.Approach. To address these issues, we propose spatio-temporal cross-attention fusion (STCAFusion), a spatiotemporal cross-attention (CA) framework that integrates EEG and EMG through multi-band based dual-branch convolutional encoders and parallel temporal and spatial CA modules. This design enables detailed modeling of inter-modal correlations across both time and space. We evaluate STCAFusion on a newly collected dataset of synchronous EEG-EMG recordings from 12 subjects, where the data were acquired under two paradigms (Reaching and Lifting) designed from daily functional upper-limb activities to emphasize directional and strength control.Main results. With leave-one-run-out cross-validation, STCAFusion achieves average accuracies of 84.15% and 95.22% in the two paradigms, outperforming the strongest competing EEG-EMG fusion baselines by 3.4% in the Reaching paradigm and 1.8% in the Lifting paradigm. Visualization of learned attention weights further reveals meaningful spatiotemporal EEG-EMG coupling patterns, offering insights into neural-muscular coordination patterns relevant to rehabilitation-oriented BCI design.Significance. These results highlight the potential of CA-based multimodal physiological signal fusion in building reliable hybrid BCI and wearable devices for upper-limb control and rehabilitation.
14 July 2026
Read appraisal →Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology
Effects of transcranial direct current stimulation on motor function in patients with multiple sclerosis: a systematic review and meta-analysis
OBJECTIVE: To assess the effectiveness of transcranial direct current stimulation (tDCS) in enhancing motor function recovery in the upper and lower extremities of individuals with multiple sclerosis. DATA SOURCES: PubMed, Embase, Web of Science, Cochrane Library, and China Knowledge Network (CNKI) databases from inception until December 28, 2025. STUDY SELECTION: Randomized controlled trials that investigated transcranial direct current stimulation and tDCS combined with physical training on motor function in the upper and lower limbs of individuals with MS. DATA EXTRACTION: Two reviewers extracted the data and assessed the risk of bias. Standardized mean difference (SMD) and 95% confidence intervals (CI) were calculated using a random-effects model based on heterogeneity. The analysis included 10 articles with a total of 299 subjects. DATA SYNTHESIS: The Meta-analysis shows tDCS significantly improves lower limb motor function in MS patients (SMD = 0.31, 95% CI: 0.16 - 0.46, p < 0.0001). Subgroup analyses found significant increases in walking distance (p = 0.0005) and static balance (Berg Balance Score, p = 0.0002), but no significant improvements were observed in walking speed and dynamic balance. For upper limb motor function, no statistically significant difference was found in hand dexterity (9-HPT) (p = 0.11). Notably, subgroup analyses of stimulation parameters showed that high intensity (2.0-2.5 mA, p = 0.007) and a 20-minute duration (p = 0.03) yielded significant effects, whereas low intensity (1.0-1.5 mA) and 15-minute duration showed no significant benefits. CONCLUSIONS: Transcranial direct current stimulation is effective in enhancing lower limb motor function in patients with MS. However, the efficacy of tDCS in improving upper limb fine motor skills remains unclear. Evidence suggests that stimulation at 2.0-2.5 mA for 20 min may be the optimal protocol for clinical intervention.
13 July 2026
Read appraisal →Journal of medical systems
Automatic Sleep Staging Using Cardiorespiratory Signals: A Systematic Review of Methodologies and Performance
Cardiorespiratory-based methods offer promising alternatives to traditional PSG for longitudinal sleep monitoring, holding significant systemic medical value for scalable sleep health management. This systematic review synthesizes methodological frameworks and performance outcomes of automatic sleep staging using cardiorespiratory signals. Four databases were searched and a total of 35 studies published since 2010 were identified. The analysis revealed that cardiorespiratory signal-based sleep staging achieved a practically meaningful accuracy of 70%, with no significant performance differences observed among signal modalities (cardiac signals, cardiorespiratory signals, or cardiac/cardiorespiratory signals combined with other non-EEG modalities) or between modeling algorithms (traditional machine learning vs. deep learning). However, we identified significant methodological heterogeneity and several critical model failure modes that hinder clinical translation, including the widespread lack of external validation, consistently poor classification of the N1 sleep stage, and limited generalization across diverse patient populations. To realize the technology's potential, future research must establish consensus-driven methodological guidelines and rigorously validate algorithms on large, demographically and clinically diverse datasets. These advances are essential for integrating cardiorespiratory-based sleep staging into healthcare systems as a scalable tool for population-level screening, longitudinal monitoring, and tiered clinical decision support.
8 July 2026
Read appraisal →Brain topography
Local Field Potential Recordings Using Deep Brain Stimulation: A Practical Workflow and Open-Source Signal Processing Pipeline
Deep brain stimulation (DBS) systems with electrophysiological recording capabilities offer a unique opportunity to chronically record local field potentials (LFP) from implanted electrodes, enabling real-time monitoring of subcortical neural dynamics and the development of adaptive neuromodulation strategies. However, translating raw LFP into clinically relevant biomarkers remains challenging due to numerous technical considerations. To address this, we share both a practical guide for conducting LFP experiments and BrainWave-DBS, an open-source pipeline that implements the workflow and supports analysis of patient-derived LFP data. The pipeline covers data import, preprocessing, spectral analysis, and visualization, offering a reproducible framework for LFP analysis. Furthermore, we demonstrate the functionality of the pipeline using a representative Parkinson's disease dataset and highlight key technical insights gained through this work. By improving access to neural signal acquisition and analysis, our goal is to foster broader adoption of this technology, which could advance closed-loop neuromodulation of deep brain structures.
29 June 2026
Read appraisal →PloS one
Hybrid deep learning for mental workload classification using EEG with enhanced preprocessing and interpretability
Mental workload classification is critical in safety-sensitive fields such as healthcare and aviation. However, electroencephalography-based approaches still face challenges with generalizability, noise robustness, and interpretability. In this study, we propose an integrated hybrid deep learning framework to address these limitations and enable robust, interpretable electroencephalography-based mental workload classification. The proposed approach uses a Variational Autoencoder to enhance noise reduction and feature extraction from band-wise topographical videos, a Convolutional Block Attention Module to adaptively focus on important spatial-channel Electroencephalogram features, and a Bidirectional Long Short-Term Memory network to capture complex temporal dependencies under leave-one-subject-out cross-validation. We conducted ablation studies to identify each architecture component's contribution and sensitivity analyses to determine the optimal parameters. The model achieved the highest overall accuracy among the baselines and reached an average accuracy of 83.9% across subjects for classifying four mental workload levels. Ablation studies confirmed the added value of all three architecture components for improving performance. Sensitivity analyses identified optimal parameters, including a 10-second window length that balances temporal context and specificity. For better interpretability and neurophysiological insight, we used Gradient-weighted Class Activation Mapping to visualize key frontal-parietal brain regions and frequency bands associated with workload dynamics. Future research could explore adaptive windowing strategies, multimodal data integration, cross-dataset benchmarking, and evaluation against transformer-based and graph neural network architectures under consistent subject-independent evaluation settings to further enhance model generalizability.
28 June 2026
Read appraisal →PloS one
Bayesian Uncertainty-aware Deep Learning with noisy labels: Tackling annotation ambiguity in EEG seizure detection
Deep learning is advancing EEG processing for automated epileptic seizure detection and onset zone localization, yet its performance relies heavily on high-quality annotated training data. However, scalp EEG is susceptible to high noise levels, which in turn leads to imprecise annotations of the seizure timing and characteristics. This "label noise" presents a significant challenge in model training and generalization. In this paper, we introduce Bayesian UncertaiNty-aware Deep Learning (BUNDL), a novel algorithm that informs a deep learning model of label ambiguities, thereby enhancing the robustness of seizure detection systems. By integrating domain knowledge into an underlying Bayesian framework, we derive a novel KL-divergence-based loss function that capitalizes on uncertainty to better learn seizure characteristics from scalp EEG. Thus, BUNDL offers a straightforward and model-agnostic method for training deep neural networks with noisy training labels that does not add any parameters to existing architectures. Additionally, we explore the impact of improved detection system on the task of automated onset zone localization. We validate BUNDL using a comprehensive simulated EEG dataset and two publicly available datasets collected by Temple University Hospital (TUH) and Boston Children's Hospital (CHB-MIT). Results show that BUNDL consistently identifies noisy labels and improves the robustness of three base models under various label noise conditions. We also conduct ablation experiments on uncertainty quantification, evaluate cross-site generalizability to Siena EEG dataset, and quantify computational cost of all methods. Furthermore, we demonstrate that BUNDL improves seizure onset zone localization accuracy. Ultimately, BUNDL presents as a reliable method that can be seamlessly integrated with existing deep models used in clinical practice, enabling the training of trustworthy models for epilepsy evaluation.
25 June 2026
Read appraisal →Journal of neural engineering
Computational neuroelectrophysiology and artificial intelligence for drug-resistant epilepsy: recent advances, current challenges, and future directions
Drug-resistant epilepsy (DRE) affects approximately 30% of epilepsy patients, with surgical cure rates below 70%. This challenge drives a fundamental paradigm shift from localizing a discrete epileptogenic zone toward characterizing and modulating the dysfunctional brain networks that initiate and propagate seizures. This review critically synthesizes how computational neuroelectrophysiology and artificial intelligence (AI) converge to propel this shift. We chart the trajectory from quantifying local biomarkers, including high-frequency oscillations and excitation-inhibition balance metrics, to mapping the topological properties of epileptic networks through functional and effective connectivity. The integration of these network features with AI techniques, particularly spatiotemporal deep learning architectures, has demonstrated significant potential for enhancing both the localization of the epileptogenic network and the prediction of postoperative outcomes. However, the translation of this network-centric paradigm into clinical practice remains constrained by several challenges, including spatial sampling bias, the inherent label instability of surgical outcomes, and the 'black-box' interpretability crisis. Future directions must emphasize EEG foundation models, causal AI, and the development of interactive multimodal AI agents. The convergence of mechanism-driven network models and data-driven AI frameworks promises a new era of personalized, network-guided therapy for DRE.
11 June 2026
Read appraisal →Brain structure & function
Automated interictal epileptic spike detection from simple and noisy annotations in MEG data
In drug-resistant epilepsy, presurgical evaluation can be considered for suitable candidates. Magnetoencephalography (MEG) has been shown to be an effective exam to inform the localization of the epileptogenic zone through the localization of interictal epileptic spikes. Manual detection of these pathological biomarkers remains a fastidious task due to the high dimensionality of MEG recordings, and inter-rater agreement has been reported to be only moderate. Current automated methods are unsuitable for clinical practice, either requiring extensively annotated data or lacking robustness on non-typical data. Considering a database of 82 patients, we demonstrate that deep learning models can be used for detecting interictal spikes in MEG recordings, even when only temporal and single-expert annotations are available, which represents real-world clinical practice. We propose two model architectures: a feature-based artificial neural network (ANN) and a convolutional neural network (CNN), evaluated against a state-of-the-art model to classify short time windows of signal. In addition, we employ an interactive machine learning strategy to iteratively improve our data annotation quality using intermediary model outputs. Both proposed models outperform the state-of-the-art model (F1-scores: CNN = 0.46, ANN = 0.44) when tested on 10 holdout test patients. The interactive machine learning strategy demonstrates that our models are robust to noisy annotations. Overall, results highlight the ability of our models to analyze complex and imperfectly annotated data. Our method of interactive machine learning offers great potential for faster data annotation, while our models represent useful and efficient tools for automated interictal spikes detection.
1 June 2026
Read appraisal →Scientific reports
From sleep staging to spindle detection: a case study on end-to-end automated sleep analysis
Automation of sleep analysis, including both macrostructural (sleep stages) and microstructural (e.g., sleep spindles) elements, promises to enable large-scale sleep studies and to reduce variance due to inter-rater incongruencies. While individual steps, such as sleep staging and spindle detection, have been studied separately, the feasibility of automating multi-step sleep analysis remains unclear. In this case study, we evaluate whether a fully automated analysis using validated machine learning models for sleep staging (RobustSleepNet) and subsequent spindle detection (SUMOv2) can replicate findings from an expert-based study of bipolar disorder. The automated analysis qualitatively reproduced key findings from the expert-based study, including significant differences in fast spindle densities between bipolar patients and healthy controls, accomplishing in minutes what previously took months to complete manually. While the results of the automated analysis differed quantitatively from the expert-based study, possibly due to biases between expert raters or between raters and the models, the models individually performed at or above inter-rater agreement for both sleep staging and spindle detection. Our results demonstrate that fully automated approaches have the potential to facilitate large-scale sleep research. We are providing public access to the tools used in our automated analysis by sharing our code and introducing SomnoBot, a privacy-preserving sleep analysis platform.
25 May 2026
Read appraisal →Studies in health technology and informatics
AI-Integrated EEG Decision Support for Neurocritical Care: A Conceptual and Feasibility Framework
The clinical use of electroencephalography (EEG) for neuro-prognostication in neurocritical care remains limited, despite its ability to provide non-invasive insight into brain activity. This study presents a conceptual framework and early-stage prototype of an AI-assisted EEG decision-support system. Using a literature-informed design and system architecture approach, a web-based prototype was developed to support automated EEG analysis and interpretable brain response reporting. Feasibility testing with four users demonstrated reliable system operation and generation of within-patient decision-support indicators. We highlight the system design and feasibility of operationalizing EEG-based analytics into a clinically interpretable decision-support framework for neurocritical care.
24 May 2026
Read appraisal →Neurosurgical review
Preoperative predictors of seizure outcomes after epilepsy surgery for mesial temporal sclerosis: A systematic review
Mesial temporal lobe epilepsy (MTLE) due to hippocampal sclerosis (HS) is a leading cause of drug-resistant epilepsy. Surgical resection, particularly anterior temporal lobectomy (ATL), achieves high rates of seizure freedom, but outcomes vary. Identifying reliable preoperative predictors may optimize patient selection and surgical success. To systematically evaluate preoperative clinical, neuroimaging, and neuropsychological factors associated with postoperative seizure outcomes in MTLE. A systematic review was conducted according to PRISMA 2020 guidelines. PubMed, Scopus, and Web of Science were searched for cohort studies (2010-2025) reporting predictors of seizure freedom after MTLE-HS surgery. Eligible studies assessed one or more preoperative factors (clinical, EEG, imaging, neuropsychological) in relation to postoperative outcomes. Data were extracted for qualitative synthesis, and methodological quality was assessed with the Newcastle-Ottawa Scale (NOS). From 615 records, 24 studies were included, encompassing patients treated with ATL, selective amygdalohippocampectomy (SAHE), or stereotactic laser ablation. Seizure freedom rates (Engel I/ILAE 1) ranged from 50% to 75%, with ATL achieving ~ 70% at 1-2 years. Multimodal concordance-MRI-visible HS, unilateral seizure onset on EEG, and concordant PET hypometabolism-was the strongest predictor of seizure freedom. Bilateral EEG/PET abnormalities consistently predicted poorer outcomes, whereas structural measures, such as hippocampal subfield volumes and HS subtypes, showed little prognostic value. Multimodal presurgical concordance is a robust predictor of favorable outcomes in MTLE-HS surgery, whereas bilateral or discordant findings indicate a higher risk of failure. Comprehensive presurgical evaluation remains essential to optimize surgical decision-making.
21 May 2026
Read appraisal →PloS one
Deep Learning outperforms physicians in myopathy and neuropathy classification based on Needle Electromyography Signal
Needle electromyography (nEMG) is a valuable tool for diagnosing patients with neuromuscular diseases. However, it is labor-intensive and is prone to diagnostic inaccuracies stemming from human biases. To address these challenges, we validated an nEMG diagnosis-aiding system with minimal preprocessing using deep learning model to classify patients into three categories: normal, myopathy, and neuropathy. Using 376 nEMG signals from 57 patients from a tertiary university hospital database through nested k-fold cross validation, deep learning model surpassed the classification performance of six electromyographers. The median patient classification accuracy, precision, sensitivity, and specificity of the deep learning model was 0.70, 0.70, 0.70, and 0.85, respectively, whereas those of the physicians were 0.55, 0.60, 0.54, and 0.78, respectively. Model interpretability and failure analysis showed that the deep learning model classifies based on relevant signal features. Despite higher accuracy of DL model, the number of unanimously misclassified cases were higher in the DL model than physicians. Our study validates deep learning is a fast, accurate, and practical application to aid physicians in diagnosing patients using nEMG signals.
20 May 2026
Read appraisal →Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology
Parkinson disease severity detection based On OPtFuzNet with fused features
BACKGROUND: Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting millions of people worldwide. It severely impairs patients' mobility. For effective treatment strategies, it is essential to determine the severity of the disease at an early stage. METHODS: In this study, an optimized feature-fusion framework is developed to detect the severity of Parkinson's disease using gait data via an Optimized Fuzzy Neural Network (OPtFuzNet). This feature-fusion framework can be categorized into three stages: pre-processing, feature extraction, and classification. In the pre-processing stage, a median filter is employed to reduce noise present in the gait images. During feature extraction, local and global features of the gait images are captured using SAE and IDCN, respectively. Subsequently, these features are fused and fed into the OPtFuzNet classifier, where parameter optimization is performed using IGWO-Lévy. RESULTS: The proposed model is evaluated on the GAIT-IT and GAIT-IST benchmark datasets using standard metrics, achieving superior performance with accuracies of 98.08% and 98.12%, respectively. Furthermore, a feature importance analysis is conducted to identify the most influential gait characteristics that contribute to determining the severity of the impairment, thereby enhancing the clinical interpretability of the model.The proposed model is evaluated on the GAIT-IT and GAIT-IST benchmark datasets using standard metrics, achieving superior performance with accuracies of 98.08% and 98.12%, respectively. Furthermore, a feature importance analysis is conducted to identify the most influential gait characteristics that contribute to determining the severity of the impairment, thereby enhancing the clinical interpretability of the model. CONCLUSION: These results validate the efficacy of the proposed feature fusion and optimization strategy in identifying discriminative gait patterns associated with the severity of Parkinson's disease. However, as the datasets are based on simulated gait patterns rather than real clinical data, further validation on real-world datasets is necessary for practical applicability.
20 May 2026
Read appraisal →Cerebellum (London, England)
Inhibitory Circuits of the Sensorimotor Network are Not Modulated By Cerebellar Transcranial Direct Current Stimulation in Isolated Cervical Dystonia
Impaired cerebellar influence on motor cortical excitability and plasticity has been reported in cervical dystonia (CD) patients, accompanied by the absence of cerebellar brain inhibition (CBI). Polarity-specific modulation of CBI in healthy individuals using cerebellar transcranial direct current stimulation (ctDCS) suggested that ctDCS could normalize abnormal cerebellar output and improve clinical symptom severity in CD. The objective of this study was to determine whether anodal or cathodal ctDCS can modulate neurophysiological parameters of cortico-cortical, cerebello-cortical or afferent inhibition and improve motor symptom severity in CD patients. Fifteen patients with isolated CD participated in a randomized, double-blinded crossover study consisting of three sessions of anodal, cathodal, or sham ctDCS. Before and after each intervention, motor symptom severity and inhibitory circuits of the sensorimotor network were investigated using transcranial magnetic stimulation (TMS), including short-interval intracortical inhibition (SICI), short-latency afferent inhibition (SAI) and CBI. Baseline TMS measurements showed inhibitory influence of SICI (p < 0.001) and SAI (p < 0.001) in CD patients, whereas CBI had no inhibitory effect (p = 0.281). The different ctDCS interventions caused no significant modulation in any of the inhibitory TMS paradigms investigated. Similarly, motor symptom severity remained unchanged after the ctDCS interventions. A single session of ctDCS was not effective to modulate inhibitory circuits of the sensorimotor network in isolated CD patients. Future studies focusing on repetitive ctDCS interventions or multifocal stimulation protocols are needed to assess the influence on network excitability and connectivity as well as on motor symptoms.
18 May 2026
Read appraisal →Neurology. Clinical practice
Optimizing Antiseizure Medication Taper in the Epilepsy Monitoring Unit
BACKGROUND AND OBJECTIVES: Inpatient video-EEG epilepsy monitoring aims to provide a safe environment for antiseizure medication (ASM) tapering to precipitate seizures, which is often required for diagnosis and treatment. However, there is no standardized taper protocol across institutions, which may contribute to increased morbidity through longer hospitalizations if the taper is too mild and bilateral tonic-clonic seizures (BTCSs) if too aggressive. This study investigates how ASM taper influences length of stay (LOS) and patient outcomes in the epilepsy monitoring unit (EMU), aiming to develop guidelines for medication taper strategies to maximize diagnostic yield while minimizing patient risk. METHODS: We used natural language processing to extract data from electronic medical records of 639 patients admitted to the University of Pennsylvania EMU between 2017 and 2023. We estimated ASM load, the total normalized sum of all ASM serum concentrations based on dose-dependent pharmacokinetic models, throughout each hospitalization. We applied LASSO (Least Absolute Shrinkage and Selection Operator) regression and mixed-effects models to identify predictors of LOS and BTCS risk as a function of taper speed and percentage of baseline ASM load in patients with (hBTCS) and without (nBTCS) a history of BTCS. RESULTS: Taper speed did not correlate with BTCS risk, although absolute ASM load did. Median ASM loads at seizure occurrence were 51% of baseline in the hBTCS group, 41% in the nBTCS group, and 39% in the unclear group. BTCS occurred at higher ASM loads in patients with hBTCS vs nBTCS. Median ASM loads at first EMU seizure were 33% for both hBTCS and nBTCS groups, while the unclear group had first seizures at lower loads (25% of baseline). Time to first seizure and number of baseline ASMs were positively correlated with longer hospital stays, suggesting that sequential medication taper delayed seizures and prolonged LOS without altering BTCS risk. DISCUSSION: Patient history and ASM load are critical factors in predicting BTCS. Rapid taper to approximately 50% baseline ASM load may more efficiently produce seizures without increased BTCS risk in patients without BTCS history. Individualized taper protocols remain advised, particularly for severe epilepsies. This study provides a foundation for quantitative, evidence-based guidelines to optimize ASM reduction strategies for safely precipitating seizures in EMUs.
15 May 2026
Read appraisal →Progress in biomedical engineering (Bristol, England)
Cross-subject generalization for EEG decoding: a survey of deep learning methods
Deep learning for cross-subject electroencephalography (EEG) decoding is hindered by high inter-subject variability, which introduces a severe domain shift between training and unseen test subjects. This survey presents a comprehensive review of deep learning methodologies specifically engineered to address this cross-subject generalization challenge. To ground this analysis, we formalize the cross-subject setting as a multi-source domain problem and delineate the rigorous, subject-independent evaluation protocols required for valid assessment. Central to this survey is a systematic taxonomy of the current literature into discrete methodological families, including feature alignment, adversarial learning, feature disentanglement, and contrastive learning. We conclude by examining three critical elements for advancing robust, real-world decoding: the theoretical limitations of current methodologies, the structural value of subject identity, and the emergence of EEG foundation models.
14 May 2026
Read appraisal →Biomedical physics & engineering express
Differentiating depression subtypes through three-electrode EEG and emotional stimuli: a machine learning approach
We propose an EEG-based framework for depression subtype assessment using emotion-modulated neural dynamics elicited by immersive virtual reality (VR). EEG was recorded from 70 participants (31 first depressive episode, FDE; 18 recurrent depressive episode, RDE; 21 control participants, HC) using a compact frontal montage (Fp1/Fpz/Fp2) during positive and negative VR conditions, focusing on the immediate post-stimulus regulation period. Subject-level features were analyzed with a linear mixed-effects model to probe group-by-condition interaction patterns, and a Bi-Emotional Siamese Network (BESN) was developed to model baseline-anchored within-subject deviations by fusing positive/negative reactivity streams with path-signature-based temporal encoding. Feature-level analyses revealed interaction signals suggestive of differential emotion-related modulation across groups. In subject-wise evaluation, BESN achieved 83.1% accuracy for HC versus FDE discrimination and 70.6% accuracy for three-class classification (HC versus FDE versus RDE), outperforming conventional machine-learning baselines. Robustness was further supported on an external public resting-state EEG dataset (MODMA), achieving 78.2% accuracy. These results suggest that baseline-anchored, emotion-modulated EEG dynamics combined with path-signature modeling provide an objective and generalizable computational approach for depression-related group discrimination.
14 May 2026
Read appraisal →Experimental brain research
Subthalamic nucleus deep brain stimulation and associations with attention and processing speed in Parkinson's disease
Although subthalamic nucleus deep brain stimulation (STN DBS) is an effective treatment for the motor symptoms of Parkinson's disease (PD), its effects on cognitive functions are controversial. This cross-sectional study aimed to compare attention and processing speed in PD patients with STN-DBS and those receiving medical therapy, using a multimodal assessment combining P300 event-related potentials and neuropsychological tests. The study included 20 PD patients who underwent STN DBS and 17 PD patients who were followed with medical treatment. Demographic and clinical data for all patients were recorded. Cognitive functions were assessed using the P300 potential, the Stroop test (ST), and the Serial Digit Learning Test (SDLT). P300 potential latency and amplitude values, as well as neuropsychometric test results, were statistically compared between the two groups. P300 latency was shorter in the DBS-treated group than in the non-DBS group (p = 0.004). Performance times on Part III of the ST were shorter in the DBS-treated group (p = 0.03). In addition, a higher proportion of patients in the DBS-treated group performed normally on Part I of the ST compared to the non-DBS group (p = 0.03). There was no significant difference between the two groups in SDLT results. Shorter P300 latency and subtle differences in ST performance suggest a potential association between STN-DBS and the cognitive domains of attention and processing speed. Combining neuropsychological and electrophysiological assessments could provide a more comprehensive evaluation, and future large-scale, longitudinal studies are needed to clarify domain-specific cognitive effects.
14 May 2026
Read appraisal →Angewandte Chemie (International ed. in English)
A Novel Acetylcholine Nanosensor for Single Vesicle Storage and Sub-Quantal Exocytosis in Living Neurons and Organoids
Acetylcholine (ACh) is a critical neurotransmitter that regulates diverse physiological functions, such as cognition and muscle contraction, through synaptic transmission. However, in situ quantitative chemical analysis of single-vesicle storage and release dynamics at single-cell level remains a major technical challenge, hindering mechanistic understanding of cholinergic synaptic plasticity across physiological and pathological states. To address this, we developed an Ti3C2Tx MXene/enzyme-functionalized (M@E@CF) nanosensor, enabling real-time monitoring of vesicular ACh storage and exocytotic release in primary cholinergic neurons and human spinal cord organoids. Our findings reveal a sub-quantal release mode in both mouse and human-derived neurons. To further elucidate the regulatory principles of exocytosis kinetics, we classified single-vesicle exocytosis patterns based on signal peak shapes using a 1D convolutional neural network (1D-CNN) deep learning model, uncovering significant differences in the number of released molecules and kinetic parameters across modes. Critically, in Down syndrome models, we observed significantly reduced single vesicle ACh storage and release alongside an elevated release fraction, concurrent with shortened fusion pore durations during exocytosis. Thus, the M@E@CF nanosensor platform establishes a versatile tool for spatiotemporal investigation of neurotransmitter storage and release dynamics, providing a critical technical foundation for exploring physiological functions and pathological mechanisms of the cholinergic system.
13 May 2026
Read appraisal →Journal of neural engineering
Uncovering relationships in multi-channel EEG data using principal Hessian directions and Ricci flow
Objective.The high-dimensional nature of multi-channel EEG data poses major challenges for downstream classification. Uncovering connectivity relationships and network structure between EEG channels can improve high-dimensional EEG signal recovery by identifying redundant inputs and guiding more targeted feature selection. In this study, we aim to demonstrate a systematic framework for extracting useful interchannel structure in EEG data.Approach.We present two complementary methods for inferring network structure and connectivity in EEG data: (1) a new supervised algorithm for inferring classification-relevant community structure based on principal Hessian directions (pHds), and (2) a discrete Ricci flow-based unsupervised community detection algorithm. We demonstrate these systematic methods on high-dimensional real-world EEG datasets involving classifying imagined digits versus non-digits and detecting emotional valence.Main results.We show that our pHd and Ricci flow methods-when combined-can detect interchannel relationships that meaningfully hold on unseen test classification data. Moreover, we demonstrate that this interchannel structure extracted via pHd and Ricci flow can enable subsequent improvements in downstream EEG signal classification.Significance.Our combined pHd-Ricci flow method expands the existing EEG preprocessing toolkit by offering a systematic framework for extracting meaningful network structure from high-dimensional EEG data. By facilitating more targeted and effective feature selection, our method has the potential to improve EEG signal recovery in real-world applications.
11 May 2026
Read appraisal →Epilepsy research
Machine learning in epilepsy
Epilepsy is a complex neurological disorder characterized by pathological processes that unfold across multiple biological scales, from cellular excitability and synaptic integration to large-scale network dynamics observable in electroencephalographic (EEG) recordings. While traditional analytical approaches have provided valuable insights, they often fail to capture high-dimensional and nonlinear structure of contemporary electrophysiological and clinical datasets. Consequently, machine learning (ML) has emerged as a powerful analytical framework in epilepsy research, although its rapid adoption has revealed a growing gap between algorithmic performance and biological interpretability. This review examines ML methods operating across three analytically distinct yet interconnected levels: (i) unsupervised learning for cellular-level phenotyping using high-dimensional electrophysiological data; (ii) supervised learning for EEG-based seizure detection and prediction; and (iii) multiscale modeling frameworks integrating neuronal and network dynamics. Rather than providing an exhaustive catalog of algorithms, we focus on inferential assumptions underlying ML applications, the methodological pitfalls constraining generalization and clinical relevance, and how ML-derived representations can be interpreted within established neurophysiological theory. We highlight that unsupervised ML facilitates identification of latent excitability phenotypes and trajectories obscured in traditional univariate analyses, while supervised ML has substantially advanced automated seizure detection and prediction, despite persistent challenges related to data leakage, class imbalance, and ambiguous preictal labeling. We argue that the most promising direction lies in embedding ML within multiscale mechanistic models, where data-driven inference facilitates parameter estimation and hypothesis generation rather than black-box prediction. By prioritizing interpretability, rigorous validation, and cross-scale integration, ML-enhanced multiscale frameworks offer a path toward clinically actionable models of epilepsy.
9 May 2026
Read appraisal →Journal of medical systems
Interpretable Machine Learning to Anticipate the Diagnostic Yield of EEG in the Emergency department. The EMINENCE study.
INTRODUCTION: Emergent electroencephalography (emEEG) is increasingly employed in the emergency department (ED) for evaluating altered consciousness and seizure-related conditions, yet standardized criteria guiding its use remain limited. METHODS: We retrospectively analyzed 1,018 patients (mean age 66 ± 20 years; 48.4% female) undergoing emEEG at the ED of the Careggi Teaching Hospital (Florence, Italy) in 2023. Clinical, anamnestic, and neuroimaging data available at admission were used to train supervised machine-learning (ML) models. We evaluated tree-based ensembles (Random Forest and XGBoost) to predict abnormal and epileptiform emEEG, as well as confirmation or refutation of initial diagnosis. Ground-truth labels were derived from a multidisciplinary expert team including neurologists, neurophysiopathologists and intensivists. Model performance was assessed with 5 × 5 nested cross-validation, receiver operating characteristic (ROC) analysis, balanced accuracy, decision-curve analysis, and Shapley Additive Explanations (SHAP) interpretability. RESULTS: Abnormal emEEG occurred in 691 cases (67.9%), epileptiform activity in 192 patients (18.9%). emEEG ruled out the initial diagnostic suspicion in 514 cases (50.5%) and confirmed it in 188 cases (18.5%). Best performance was obtained with Random Forest for abnormal emEEG (AUC 0.79, 95% CI: 0.76-0.82) and diagnosis rule-out (0.84, 0.81-0.86), and with XGBoost for epileptiform emEEG (0.82, 0.78-0.85) and diagnosis confirmation (0.82, 0.79-0.85). Performance varied by initial diagnostic suspicion, but subgroup-stratified analyses showed overall consistent patterns. Key predictive features included altered consciousness, prior brain lesions, antiseizure therapy, and seizure-related presentations. Interpretability analyses revealed seizure-centric features drove confirmation, while systemic or nonspecific features favored refutation. CONCLUSIONS: Interpretable ML models using only admission data can predict emEEG outcomes and anticipate their diagnostic contribution, supporting triage and decision-making in emergency neurology without replacing clinical judgment. Models and explanations were easily usable on a freely-accessible website ( www.emergencyeeg.com ), where tools return probabilistic outputs for all four prediction tasks together with per-patient explanation plots, enabling transparent and reproducible use. CLINICAL TRIAL NUMBER: Not applicable.
9 May 2026
Read appraisal →Pediatric neurology
Electroencephalography-Based Machine Learning Models for Predicting Ketogenic Diet Outcomes in Pediatric Drug-Resistant Epilepsy
BACKGROUND: Ketogenic diet therapy (KDT) is an established treatment for drug-resistant epilepsy (DRE); however, methods for predicting its effectiveness remain underdeveloped. This study evaluated various machine learning (ML) models in predicting responses to KDT among DRE patients based on electroencephalography data. METHODS: Using leave-one-out cross-validation, this study evaluated 19 ML classifiers in predicting the outcomes of 90 DRE patients based on absolute and relative power, as well as functional connectivity measures (phase-locking value, phase lag index [PLI], and weighted PLI) across standard frequency bands. KDT significantly reduced seizure frequency at three and 6 months after initiation. RESULTS: The most effective classifier at 3 months was a Coarse Tree classifier trained on absolute power (recall = 0.933, precision = 0.767, F2 = 0.894, area under the receiver operating characteristic curve = 0.607). The most effective classifier at 6 months was a Gaussian Naive Bayes classifier trained on weighted PLI + relative power (recall = 0.759, precision = 0.854, F2 = 0.776, area under the receiver operating characteristic curve = 0.603). CONCLUSIONS: This study identified the most effective ML models for predicting KDT outcomes in DRE patients. The results highlight the potential of electroencephalography-based ML tools for guiding KDT treatment in clinical practice.
9 May 2026
Read appraisal →International journal of neural systems
Data-related Ablation for Reinforcing Deep Learning in Explaining Complex Phenomena
Deep Learning (DL) models excel at automatically learning intricate patterns within complex data, but their black box nature undermines human trust. To address this, current validation strategies typically focus on the model itself, modifying its architecture to assess the role and importance of the components. However, this model-centric view overlooks the critical learning substrate, which is represented by the data, implicitly assuming that it accurately represents the target phenomenon. This implicit trust in data means that evaluation may fail to detect whether high performance stems from exploiting biases or data quirks rather than learning relevant patterns. We present a novel data-related ablation as a complement to the traditional architectural ablation. Using this framework for Electroencephalography (EEG) signals of Emotional Recognition (ER) and Motor Execution (ME) as a case study, we show that seemingly high-accuracy models often rely heavily on process-irrelevant features, maintaining performance even when key information is eliminated. This shows that a standard, data-independent evaluation can be misleading about whether a model truly captured the intended process; the proposed approach helps distinguish robust learning from leaning on incidental characteristics. Therefore, incorporating data-related ablation is essential for developing reliable and generalizable DL models in fields that rely on data derived from complex and often not completely known phenomena.
4 May 2026
Read appraisal →Biomedical physics & engineering express
A novel machine learning approach for prediction of postoperative delirium using multi-domain features of high-density EEG
Background.Postoperative delirium (POD) poses a significant risk to patients, and accurate prediction of POD can provide guidance for positive interventions. Although many studies have applied machine learning (ML) to electronic health records to predict POD, there has been a lack of studies utilizing electroencephalogram (EEG) data to accurately predict POD.Methods.This retrospective study included 142 patients from two hospitals, among whom 33 developed POD. We extracted multi-domain features from high-density EEG (HD-EEG), screened 39 key features using recursive feature elimination, and analyzed the feature space structure using linear discriminant analysis (LDA). Four ML models, namely support vector machine (SVM), logistic regression (LR), random forest (RF), and gradient boosting machine, were compared. The dataset was divided using 5-fold cross-validation for model training and testing.Results.Compared to the other three models, SVM (linear kernel) performed best in predicting POD, achieving a classification accuracy of 95.71%. Alpha (8-12 Hz) and theta (4-8 Hz) powers, combined with nonlinear dynamics, critically contributed to the model. Furthermore, the performance of the linear models (SVM and LR) was significantly (p< 0.001) better than that of the nonlinear models, with LDA confirming strong feature space separability (Jensen-Shannon divergence = 0.69).Conclusions.HD-EEG data can be used to establish high-quality ML models for POD predictions. Multidomain features enable more comprehensive integration of EEG information related to delirium. The feature space structure significantly affects the performance of the prediction model, with the linear model offering distinct advantages in predicting POD.
2 May 2026
Read appraisal →PloS one
RMETNet: A cross-subject motor imagery EEG signal classification model based on TSLANet and riemannian geometry features
Motor imagery electroencephalogram (MI-EEG) analysis is essential for natural interaction and autonomous control in brain-computer interfaces (BCIs). However, deep learning models often struggle with inter-subject variability, which limits their ability to generalize across subjects. This study proposes RMETNet, a novel framework that integrates TSLANet, a spatio-temporal convolution module, and a multi-scale Riemannian geometry feature module. TSLANet suppresses noise and captures complex temporal patterns for preliminary signal decoding, while the spatio-temporal convolution module extracts higher-order representations. The Riemannian branch learns geometry-based distribution features across subjects, and the fused features are used for classification. To address inter-subject distribution shifts, RMETNet incorporates Maximum Mean Discrepancy (MMD) loss for domain adaptation, aligning feature distributions between source and target domains. Experiments show that on the four-class BCI Competition IV 2a (BCICIV2a) dataset, RMETNet achieved accuracies of 71.39% in the cross-subject setting and 80.71% in the subject-dependent setting; on the two-class BCI Competition IV 2b (BCICIV2b) dataset, it achieved 80.93% and 86.76%, respectively. The model consistently outperformed baseline algorithms. Ablation and visualization analyses further validated its effectiveness in reducing inter-subject feature distribution disparities and enhancing MI-EEG decoding. The code is available at: https://github.com/rokanfeermecer486/RMETNet.
26 Apr 2026
Read appraisal →Journal of neural engineering
Physiologically inspired modeling of cortical dynamics through spiking neural networks
Objective. The characterization of neural activity underlying neurophysiological function presents a major challenge in computational neuroscience. Several methods have been proposed to investigate cortical network dynamics by reconstructing underlying neural activity from electroencephalography (EEG) signals. However, these methods generally pose significant mathematical challenges.Approach. This study introduces a novel framework to model the underlying brain activity network from a functional and physiologically-inspired perspective, combining spiking neural networks with EEG signal analysis. The dynamics of single neurons are described by the well-known Izhikevich model, and distinct populations of cortical inhibitory and excitatory neurons are employed to model experimental EEG recordings. Functional interactions among distinct populations are mathematically formalized through connective probabilities.Main results. The proposed framework is validated by testing it on synthetic data, as well as on two experimental datasets comprising data from 30 healthy subjects undergoing a cold-pressure test (CPT), and 36 subjects undergoing a mental arithmetic stressor. Experimental results suggest that the proposed framework provides novel and complementary insights into characterizing neuronal changes in comparison to standard EEG power analysis.Significance. The proposed framework constitutes a promising tool for functionally characterizing the underlying cortical dynamics under pathophysiological conditions.
20 Apr 2026
Read appraisal →Sleep
Optimizing automated sleep stage scoring of 5-s mini-epochs: a transfer learning study
STUDY OBJECTIVES: Conventional sleep staging relies on 30-s epochs, potentially concealing transient sleep stage intrusion and reducing precision. Building on our previous study of mini-epochs, we investigated whether U-Sleep, an existing automatic deep learning-based sleep staging model with high performance in epochs, could be optimized to similar performance level in 5-s mini-epoch scoring, thereby enabling more detailed sleep characterization. METHODS: We created a dataset of 48 000 human-scored 5-s mini-epochs from 100 polysomnographies. We compared mini-epochs to human-scored epochs before U-Sleep was optimized using transfer learning and evaluated on a test set. Model performance was assessed using F1-scores, confusion matrices, stage distributions and transition rates comparing scorings of the original U-Sleep before, and the optimized U-Sleep after transfer learning to human-scored mini-epochs. RESULTS: Compared to human-scored epochs, human-scored mini-epochs captured significantly more transitions (1.70/min vs. 0.21/min, p < .001), and significantly more wake (8.4 per cent vs. 5.4 per cent), N1 (7.2 per cent vs. 5.4 per cent), and N2 (51.8 per cent vs. 40.9 per cent), less N3 (15.4 per cent vs. 25.2 per cent), and REM sleep (16.7 per cent vs. 23.0 per cent) (all p < .001). Optimizing U-Sleep improved its performance significantly from F1 = 0.74 to F1 = 0.81 (p < .05) and gave increased transition rates in the test set (original U-Sleep: 1.06/min, optimized U-Sleep: 1.34/min, human-scored mini-epochs: 1.70/min). Stage distributions did not differ between optimized U-Sleep's scorings and human-scored mini-epochs. CONCLUSION: After optimization, U-Sleep performance in mini-epochs matched the high performance levels previously reported in both human and automated 30-s epoch scoring. This demonstrates the feasibility of precise, automated high-resolution sleep staging. Future work should include external validation and application to full-night recordings. Statement of Significance Conventional 30-s epochs limit temporal resolution in sleep staging and may conceal transient intrusions of wake or sleep stages. However, no validated methods are available for high-resolution scoring. In this study, we trained and validated the state-of-the-art deep learning model U-Sleep for accurate automatic 5-s mini-epoch scoring using a large dataset of human-scored mini-epochs. The optimized model achieved a high performance, matching levels from previously reported automatic and human epoch scoring. Compared to epoch scoring, mini-epochs captured significantly more stage transitions, supporting their ability to uncover sleep dynamics that are otherwise lost. Our findings show the potential of high-resolution sleep staging for more detailed characterization of sleep architecture and demonstrate the feasibility of precise, automatic mini-epoch scoring.
19 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 →Computers in biology and medicine
Forecasting-based biomedical time-series data synthesis for open data and robust AI
The limited data availability due to strict privacy regulations and significant resource demands severely constrains biomedical time-series AI development, which creates a critical gap between data requirements and accessibility. Synthetic data generation presents a promising solution by producing artificial datasets that maintain the statistical properties of real biomedical time-series data without compromising patient confidentiality. While GANs, VAEs, and diffusion models capture global data distributions, forecasting models offer inductive biases tailored for sequential dynamics. We propose a framework for synthetic biomedical time-series data generation based on recent forecasting models that accurately replicates complex electrophysiological signals such as EEG and EMG with high fidelity. These synthetic datasets can be freely shared for open AI development and consistently improve downstream model performance. Numerical results on sleep-stage classification show up to a 3.71% performance gain with augmentation and a 91.00% synthetic-only accuracy that surpasses the real-data-only baseline.
16 Apr 2026
Read appraisal →Biomedical physics & engineering express
EEG-CMT: spatial-temporal representation of EEG for emotion recognition using convolutional neural networks and vision transformers
Background. Recent researches on electroencephalogram (EEG) based emotion recognition face challenges in effectively mapping the spatial positional relationships of EEG acquisition electrodes. Additionally, conventional models struggled to simultaneously capture both fine-grained temporal-spatial features and long-range dependencies in EEG signals.New method. To address these limitations, we propose a novel EEG data processing method that incorporates spatial relative position encoding and a hybrid neural architecture integrating convolutional neural networks (CNNs) with self-attention mechanisms. This approach systematically encodes the spatial topology of electrodes to enhance the representation of temporal-spatial information. CNNs are employed to extract localized temporal-spatial micro-patterns, while self-attention modules model global contextual dependencies across extended sequences, thereby enhancing model's representational capacity.Results. The experimental results and feature visualizations demonstrate that our method achieves state-of-the-art (SOTA) subject-dependent performance on two benchmark emotion recognition datasets, reaching an average accuracy of 97.51% on the SEED dataset and 96.13% on the SEED-IV dataset. Moreover, the learned spatial features align well with known neuroscientific patterns of emotional processing. Comprehensive ablation studies further validate the necessity and effectiveness of both the spatial-encoded data processing strategy and the hybrid architecture design.Comparison with Existing Methods. Compared to other hybrid neural network models, our proposed method (EEG-CMT) achieves the highest classification accuracy. Specifically, it outperforms baseline algorithms by margins ranging from 0.86% to 11.43% on the SEED dataset, and from 9.49% to 39.52% on the SEED-IV dataset.Conclusions. The proposed method effectively addresses key limitations in existing EEG-based emotion recognition models by jointly leveraging spatial topology and hybrid modeling techniques. These innovations significantly improve the model's ability to recognize emotions from EEG data and provide neural interpretable insights, offering a promising direction for future research in affective brain-computer interfaces.
15 Apr 2026
Read appraisal →Journal of neural engineering
EDAPT: towards calibration-free BCIs with continual online adaptation
Objective.Brain-computer interfaces (BCIs) suffer from accuracy degradation as neural signals drift over time and vary across users, requiring frequent recalibration that limits practical deployment. Our goal is to develop a framework that eliminates the need for separate calibration phases by enabling continual, real-time model adaptation to new users and changing signal characteristics.Approach.We propose EDAPT, a task- and model-agnostic framework for continual online learning. EDAPT first establishes a robust baseline decoder through population-level pretraining (PRE) on data from multiple users. It then personalizes this model during deployment using supervised continual finetuning on a trial-by-trial basis. Due to its modular design, EDAPT can be composed with unsupervised domain adaptation (UDA) techniques to further address distribution shifts.Main results.We validate EDAPT across nine datasets, three BCI paradigms, and four deep learning architectures. EDAPT consistently improves decoding accuracy over static models for nearly all subjects and datasets, raising mean balanced accuracy from 0.80 to 0.87 on representative datasets (table3). Ablation studies confirm that the combination of PRE and online finetuning is the primary driver of this performance gain, with further improvements on some datasets when using UDA techniques. We demonstrate real-time feasibility of the framework, with adaptation latencies under 200 ms on consumer-grade hardware. Our scaling analysis further reveals that decoding accuracy is primarily determined by the total pretraining data budget, rather than its specific allocation between subjects and trials.Significance.These findings demonstrate that continual online learning is a practical and effective strategy for creating high-performance, user-adaptive BCIs. By systematically addressing the bottleneck of model recalibration, EDAPT reduces a major barrier to the widespread adoption of BCI technology and helps advance neurotechnology toward robust, user-friendly, real-world applications.
8 Apr 2026
Read appraisal →Journal of neurology
Accuracy of muscle ultrasonography in detecting fasciculations for the diagnosis of amyotrophic lateral sclerosis: a systematic review and meta-analysis
PURPOSE: This systematic review and meta-analysis aims to evaluate the diagnostic accuracy of muscle ultrasonography in detecting fasciculations for the diagnosis of amyotrophic lateral sclerosis (ALS). METHODS: Following PRISMA-DTA guidelines, we systematically searched PubMed, Embase, Cochrane Library, Ovid Medline, Sinomed, Web of Science, CNKI and VIP for studies published up to July 8, 2025 that evaluated muscle ultrasonography to detect fasciculations for ALS diagnosis. The study protocol was registered in PROSPERO (CRD420251057866). Studies were screened using predefined inclusion and exclusion criteria and data were extracted. Risk of bias was assessed with QUADAS-2. Statistical analyses (Stata 16.0 and R 4.5.1 with the "midas," "metandi," and "mada" packages) were used to calculate pooled sensitivity (Sen), specificity (Spe), positive likelihood ratio (LR+), negative likelihood ratio (LR-), and diagnostic odds ratio (DOR). We constructed forest plots, hierarchical summary receiver operating characteristic (HSROC) curves, summary ROC (SROC) curves and calculated the area under the SROC curve (AUC). Univariate meta-regression and subgroup analyses explored sources of heterogeneity. Publication bias was assessed using Deeks' funnel plot asymmetry test. Fagan nomograms were also used to illustrate the changes from pre-test to post-test probability and to enhance clinical interpretability. RESULTS: Thirteen studies involving 1176 participants met the inclusion criteria. Muscle ultrasonography for fasciculation detection in ALS yielded a pooled sensitivity of 0.87 (95% CI 0.83-0.91) and specificity of 0.91 (95% CI 0.86-0.94). The pooled LR+ was 9.81 (95% CI 6.25-15.40) and LR- was 0.14 (95% CI 0.10-0.19), with a DOR of 70.03 (95% CI 41.72-117.56). The area under the SROC curve was 0.94 (95% CI 0.91-0.95). Meta-regression identified scan duration as a primary factor influencing diagnostic accuracy, with scan durations ≥ 30 s associated with higher sensitivity but relatively lower specificity. Deeks' funnel plot showed no significant asymmetry (p = 0.61), indicating no notable publication bias. Fagan nomograms showed that, at a pre-test probability of 30%, the post-test probability increased to 81% after a positive MUS result and decreased to 6% after a negative result. CONCLUSION: Muscle ultrasonography demonstrates good pooled diagnostic accuracy for detecting fasciculations in ALS and may serve as a useful adjunct to electrodiagnostic evaluation. Scan duration appears to significantly affect the diagnostic performance, with longer scanning improving sensitivity at the cost of reduced specificity. We speculate that prolonged scanning may be more useful in clinical scenarios where fasciculations are subtle or atypical, whereas shorter scanning may be sufficient when fasciculations are already readily apparent. Nevertheless, further large-scale prospective studies are needed to validate standardized scanning protocols and to better define the clinical role of MUS in ALS diagnostic pathways.
7 Apr 2026
Read appraisal →