Frontiers in Neuroinformatics最新文献

{"title":"Safety of lecanemab in real-world patients with Alzheimer's disease: a prospective, observational case series supported by FAERS signal analysis.","authors":"Haixia Ji, Yang Yang, Xingqi Wu, Erhao Bao","doi":"10.3389/fninf.2026.1977226","DOIUrl":"10.3389/fninf.2026.1977226","url":null,"abstract":"<p><strong>Objective: </strong>Lecanemab, a humanized immunoglobulin monoclonal antibody, has been approved for the treatment of Alzheimer's disease (AD). Although clinical trials have demonstrated its favorable safety profile, real-world studies are crucial for identifying rare adverse events (AEs) and evaluating post-marketing safety. This study aimed to evaluate the safety of lecanemab by combining a clinical prospective, observational case cohort with data from the FAERS database, thereby providing a robust basis for the identification of potential safety signals.</p><p><strong>Methods: </strong>A prospective and observational study was conducted on patients with mild-to-moderate AD receiving lecanemab treatment at a tertiary medical center in China from July 2024 to February 2026. Extracted data included demographic characteristics, time to onset of AEs, and clinical symptoms. Exposure-adjusted incidence rates were calculated per 100 patient-years (PYs). Disproportionality analysis was performed using the US Food and Drug Administration Adverse Event Reporting System (FAERS), and safety signals for AEs observed in the cohort were evaluated utilizing reporting odds ratios (RORs) and proportional reporting ratios (PRRs). Additionally, data from these two real-world sources were compared and cross-validated.</p><p><strong>Results: </strong>The single-center prospective study enrolled a total of 91 patients. The median duration of lecanemab treatment was 185 days. Thirteen AEs were recorded (56.48/100 PYs), with the most frequent being amyloid-related imaging abnormalities-hemosiderin deposition (ARIA-H; defined as microhemorrhages and hemosiderin deposition) (12.39/100 PYs) and infusion-related reactions (IRRs) (8.85/100 PYs). In the FAERS database, the most frequently reported preferred terms (PTs) included headache, amyloid-related imaging abnormalities-edema/effusion (ARIA-E), ARIA-H, and IRRs. The majority of AEs occurred within the first 1-2 months after initiating lecanemab treatment (<i>n</i> = 393, 45.65%). Furthermore, FAERS analysis revealed a potential safety signal for cerebral infarction. However, this finding must be interpreted cautiously due to the lack of complete data on traditional stroke risk factors and concomitant antithrombotic use in spontaneous reports.</p><p><strong>Conclusion: </strong>In this single-center prospective observational study, lecanemab demonstrated a generally favorable safety profile. Patients with AD aged 65 to 85 years represent a high-risk demographic for AEs. Particular clinical vigilance is required regarding the occurrence of AEs within the first two months of lecanemab initiation. Notably, ARIA (including both ARIA-E and ARIA-H) can occur at any phase during the treatment course.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1977226"},"PeriodicalIF":2.9,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13672516/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149215525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From pose estimation accuracy to motor biomarker reliability: a reliability-centred evaluation framework for markerless movement analysis.","authors":"Valentin Fernandez, Khabiba Abdiyeva, Sébastien Brot, Luis Perreira de Almeida, Afsaneh Gaillard, Christine Fernandez-Maloigne","doi":"10.3389/fninf.2026.1790241","DOIUrl":"10.3389/fninf.2026.1790241","url":null,"abstract":"<p><p>Markerless pose estimation has become an essential tool for quantitative movement analysis in neuroscience. However, current evaluation frameworks primarily rely on frame-wise spatial accuracy metrics, which provide only limited information about the reliability of the downstream kinematic biomarkers derived from reconstructed trajectories. In this study, we propose a reliability-centred evaluation framework designed to assess the preservation of downstream kinematic information in pose estimation pipelines. The framework was evaluated exclusively using DeepLabCut-based architectures trained and tested under identical experimental conditions. Forelimb trajectories of mice performing the Montoya staircase task were analysed using nine DeepLabCut-based convolutional neural network architectures. We introduce the Motor Biomarker Reliability Composite Score (MBRCS), which integrates four complementary dimensions of trajectory reliability: spatial accuracy, temporal stability, occlusion robustness, and kinematic fidelity. Each component was normalized to a 0-1 scale, where higher values indicate greater trajectory reliability. Although several architectures achieved comparable frame-wise localisation performance, larger differences were observed in temporal stability, occlusion robustness and downstream kinematic fidelity. Across the evaluated architectures, MBRCS values spanned almost the entire normalized range (0.00-0.99). These differences were reflected in the Normalized Jerk Score (NJS), which was used as the downstream kinematic biomarker in this study. Overall, our findings suggest that frame-wise localisation accuracy alone may not fully capture differences in trajectory reliability that become apparent when temporal stability, occlusion robustness and kinematic fidelity are jointly considered. The proposed framework provides a structured and reproducible methodology for benchmarking pose estimation pipelines according to the reliability of downstream motor biomarkers.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1790241"},"PeriodicalIF":2.9,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13670143/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149215470","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CLSP-REQA: a real-time quality-aware closed-loop seizure prediction framework with Mamba-BiLSTM and confidence-gated intervention.","authors":"Mufeng Chen, Qi Wu, Bingchao Huang, Xiwen Lai, Zekai Chen, Xinge Ouyang, Quansheng Ren","doi":"10.3389/fninf.2026.1888743","DOIUrl":"10.3389/fninf.2026.1888743","url":null,"abstract":"<p><strong>Introduction: </strong>Reliable seizure prediction is a prerequisite for closed-loop neurostimulation therapy, yet existing methods rarely account for the variability in EEG signal quality encountered in real-world deployment, and the overwhelming majority adopt non-strict evaluation protocols that overestimate generalization performance.</p><p><strong>Methods: </strong>We propose CLSP-REQA (Closed-Loop Seizure Prediction with Real-time EEG Quality Assessment), a unified framework that embeds a lightweight signal quality estimator directly within the prediction pipeline. A Real-time EEG Quality Assessment (REQA) module runs in parallel with a Mamba-BiLSTM backbone, producing a scalar quality score <i>q</i> ∈ [0, 1] that modulates output confidence through a tiered non-linear fusion function (ECLO). We conducted dedicated experiments testing whether this learned score reflects genuine signal quality rather than a redundant proxy for the prediction task, and a sensitivity analysis of the fusion thresholds against standard calibration baselines. The framework was evaluated under strict cross-patient evaluation on the CHB-MIT Scalp EEG Database (<i>n</i> = 23 subjects, 198 seizures) and the SIENA Scalp EEG Database (<i>n</i> = 14 subjects, 47 seizures).</p><p><strong>Results: </strong>On CHB-MIT, CLSP-REQA achieved an AUC-ROC of 0.7426 ± 0.0199, exceeding the unadapted cross-patient baseline of 0.69 reported by Jemal et al., using only 16 EEG channels compared to 23 in prior work, and without requiring any target-patient data or domain adaptation. On SIENA, CLSP-REQA achieved an AUC of 0.7012 ± 0.0249, exceeding the best domain-adapted cross-patient result of 0.61 on the same dataset under a strictly harder information constraint, since domain adaptation methods are given access to unlabelled target-patient data that CLSP-REQA does not use.</p><p><strong>Discussion: </strong>The framework outputs a structured four-tuple 〈<i>p, q, c</i>, Φ_SHAP〉 designed for compatibility with closed-loop neurostimulator interfaces. As both evaluation datasets are retrospective recordings from controlled hospital settings, we report this study as a feasibility demonstration rather than evidence of prospective, real-world clinical readiness, and the comparison with domain adaptation should not be read as unconditional superiority.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1888743"},"PeriodicalIF":2.9,"publicationDate":"2026-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13661927/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149180365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Window leakage in clinical EEG machine learning: matched subject-independent evaluation across ADHD and Parkinson cohorts.","authors":"Burak Aggul","doi":"10.3389/fninf.2026.1930795","DOIUrl":"https://doi.org/10.3389/fninf.2026.1930795","url":null,"abstract":"<p><p>Windowed electroencephalography (EEG) pipelines can produce misleadingly high performance when segments from the same participant occur in both training and testing. We define this contamination as <i>window leakage</i> and test it in two clinically and technically distinct public cohorts: 121 children (61 attention-deficit/hyperactivity disorder [ADHD], 60 controls) and 149 adults (100 Parkinson's disease, 49 controls). Signals were harmonized to 18 common channels and 128 Hz. All primary models used stratified 8-fold participant-disjoint evaluation, 50%-overlapping windows, participant-level probability aggregation, and validation participants drawn exclusively from each outer-training fold. A deliberately contaminated Random Forest audit contrasted random-window and participant-disjoint splitting. Random-window balanced accuracy reached 95.59% in ADHD and 92.02% in Parkinson's, but fell to 60.98% and 56.17%, respectively, when participants were held out. Thus, the inflation replicated across disorders, ages, recording tasks, and acquisition systems. Under correct evaluation, CNN-LSTM achieved the highest ADHD subject-level balanced accuracy (73.21%), whereas EEGNet achieved the highest Parkinson's balanced accuracy (63.75%). No architecture comparison remained significant after Bonferroni correction. One-, 2-, 4-, and 8-s windows produced no consistent cross-cohort ranking. Removing ATCNN-SA components also yielded dataset-dependent changes, and no ablation difference was significant after correction. Participant-bootstrap intervals remained wide despite tens of thousands of windows, indicating that inferential precision was governed by participant count rather than segment count. Window-level random splits contaminate clinical EEG performance estimates and must not be used to claim generalization to unseen individuals. Participants should be separated before model fitting, results should be reported at subject level, and split, windowing, aggregation, and uncertainty procedures should be disclosed explicitly.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1930795"},"PeriodicalIF":2.9,"publicationDate":"2026-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13658158/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149180408","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Irreversible functional network deformation induced by preprocessing scheme order in multivariate EEG recordings.","authors":"J Daniel Arzate-Mena, Paola V Olguín-Rodríguez, Maria Corsi-Cabrera, Erik A López-Alvarado, Markus F Müller","doi":"10.3389/fninf.2026.1907920","DOIUrl":"10.3389/fninf.2026.1907920","url":null,"abstract":"<p><strong>Introduction: </strong>The analysis of Electroencephalographic recordings is non-trivial, the order in which filtering and re-referencing are applied during multivariate signal preprocessing has a critical and underappreciated consequence for functional network estimation.</p><p><strong>Methods: </strong>Using EEG sleep recordings from healthy subjects, synthetic data from controlled numerical models (Nf-Tori and IAAFT surrogates), and surrogate-based statistical testing, we provide a quantitative comparison of two preprocessing schemes that differ solely in the order of filtering and re-referencing (filtering followed by re-referencing-FR, and re-referencing followed by filtering-RF).</p><p><strong>Results: </strong>We show that the RF scheme induces spurious cross-correlations that increase systematically with the frequency band and are strongly modulated by the power spectral exponent β of the signal. Critically, for signals with steep power spectra (β ≥2), a regime that characterizes slow-wave sleep, epileptic post-ictal periods, anesthesia, neonatal, and patients in vegetative states, the RF scheme produces artificial correlations that are irreversible. They cannot be eliminated <i>post-hoc</i> even when surrogate-corrected cross-correlations (SAC-matrix) are applied. The mechanism underlying this effect is due to the high-amplitude slow-frequency components that dominate the median reference signal over extended time intervals, embedding redundant information across all channels. In contrast, the FR scheme introduces only small, systematic negative shifts in cross-correlations that are not statistically significant and can be effectively corrected using surrogate-based methods.</p><p><strong>Conclusions: </strong>We therefore recommend the FR scheme combined with surrogate-corrected cross-correlations for all applications involving signals with β ≥2. Although this finding is demonstrated in EEG, it applies to any multivariate recording where spatial re-referencing and band-pass filtering are applied sequentially.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1907920"},"PeriodicalIF":2.9,"publicationDate":"2026-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13655006/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149164584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"NeuroFAIR curator: an AI-assisted framework for metadata enrichment, quality screening, curation prioritization, and FAIR readiness of brain tumor MRI data.","authors":"Faizan Ullah, Zaheer Abbas, Mohmmad Abrar, Farhan Amin, N Jon Shah","doi":"10.3389/fninf.2026.1907796","DOIUrl":"https://doi.org/10.3389/fninf.2026.1907796","url":null,"abstract":"<p><strong>Purpose: </strong>Public brain tumor MRI datasets are widely reused for classification and segmentation, but their metadata structure, technical annotation consistency, image quality, and FAIR readiness are seldom evaluated within a unified workflow. This study proposes NeuroFAIR Curator, an AI-assisted framework for metadata enrichment, computational quality screening, sample prioritization, and FAIR readiness assessment. The framework is intended for data stewardship and does not determine the clinical correctness of tumor annotations.</p><p><strong>Methods: </strong>NeuroFAIR Curator was evaluated on BRISC 2025, a two-dimensional contrast-enhanced T1-weighted MRI dataset distributed as JPEG images and PNG masks. A unified curation manifest combined images, masks, labels, split information, inferred anatomical planes, sequence descriptors, technical consistency checks, and quality indicators. A ResCNNClassifier generated label-confidence signals, while a UNet generated segmentation-derived uncertainty and agreement measures. Grad-CAM with deletion area under the curve (DAUC) supported classifier interpretation. Class- and plane-specific 1.5 × IQR limits were used for exploratory mask-area morphology screening. Seven normalized curation signals were fused into a Curation Uncertainty Index (CUI), and FAIR readiness was evaluated using an 18-sub-criterion rubric.</p><p><strong>Results: </strong>The framework generated 6,000 sample-level records linked with 15,586 external manifest entries and processed 4,793 image-mask pairs. Classification achieved 98.30% accuracy, 98.53% macro F1, and 99.79% macro AUC. Segmentation achieved Dice 82.30%, IoU 72.69%, precision 83.64%, recall 84.61%, Hausdorff distance 7.84 pixels, and HD95 4.11 pixels at a validation-selected threshold of 0.95. Exploratory morphology screening flagged 181 of 4,793 masks (3.78%) as statistically atypical; these flags were not interpreted as confirmed annotation errors. The mean Grad-CAM DAUC was 0.2996. CUI ranking captured 49 flagged samples at a 1% review budget. Findability improved from 87.50 to 100.00 and Interoperability from 36.00 to 91.98, while Accessibility and Reusability remained unchanged under the declared rubric.</p><p><strong>Conclusion: </strong>NeuroFAIR Curator converts a public brain tumor MRI corpus into a metadata-enriched, technically checked, quality-screened, review-prioritized, and FAIR-aligned neuroinformatics resource. Its morphology flags are exploratory computational signals for subsequent assessment rather than clinical judgments of mask correctness.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1907796"},"PeriodicalIF":2.9,"publicationDate":"2026-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13623914/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149106091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Montage-guided fusion network for long-term scalp electroencephalography seizure detection.","authors":"Hongtao Yin, Xin Yin","doi":"10.3389/fninf.2026.1942480","DOIUrl":"10.3389/fninf.2026.1942480","url":null,"abstract":"<p><strong>Introduction: </strong>Automatic seizure detection from long-term scalp electroencephalography (EEG) can reduce the burden of manual EEG review. However, multi-derivation EEG is commonly processed as a collection of input channels without explicitly modeling the structured relations defined by the bipolar montage.</p><p><strong>Methods: </strong>In this study, we propose MGF-Net, a montage-guided fusion network for seizure detection from long-term scalp EEG. MGF-Net represents 18 fixed bipolar derivations as nodes in a relation-labeled graph. The graph encodes same-chain adjacency, left-right correspondence, and shared-electrode relations derived from the bipolar montage. After epoch-wise robust normalization and wavelet-based signal reconstruction are applied independently to each epoch, a shared one-dimensional convolutional neural network extracts temporal features from each derivation. A relation-aware graph Transformer then performs feature fusion across derivations using relation-specific key and value projections. Attention-based graph pooling generates an epoch representation for ictal probability estimation, and post-processing converts epoch-wise probabilities into seizure-event detections.</p><p><strong>Results: </strong>Experiments on the CHB-MIT dataset show that MGF-Net achieves a segment-level accuracy of 98.38%, sensitivity of 93.18%, and specificity of 98.41%. At the event level, the proposed method detected all 65 annotated test events, achieving a sensitivity of 100.00%, a false detection rate of 0.82/h, and a latency of 2.03 s.</p><p><strong>Discussion: </strong>These results demonstrate MGF-Net's effectiveness for seizure detection from long-term scalp EEG.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1942480"},"PeriodicalIF":2.9,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13574979/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148975926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Toward FIAP^®-Digital: an interpretable multimodal AI architecture for translational stratification and precision care in Autism.","authors":"Yves Fuamba","doi":"10.3389/fninf.2026.1868068","DOIUrl":"10.3389/fninf.2026.1868068","url":null,"abstract":"<p><p>Autism spectrum disorder is characterized by substantial heterogeneity across biological burden, adaptive reserve, developmental timing, therapeutic engagement, and intervention responsiveness. This heterogeneity increasingly motivates the development of multimodal and artificial intelligence-supported approaches in autism research. However, many existing approaches emphasize diagnostic classification, symptom prediction, or generic outcome modeling, while fewer frameworks explicitly preserve clinically interpretable constructs linking biological burden, developmental state, therapeutic accessibility, and responsiveness to intervention. This Hypothesis and Theory article proposes FIAP^®-Digital as a conceptual, hypothesis-generating, and human-supervised multimodal AI architecture for future translational stratification research in autism. FIAP^®-Digital is designed to integrate multimodal inputs-including biological, physiological, developmental, contextual, therapeutic-process, and longitudinal indicators-into structured estimates of FIAP^® constructs such as the Biological Burden Index, energetic capacity, adaptive neurodevelopmental window accessibility, therapeutic engagement accessibility, neuroplastic accessibility, and higher-order translational profiles. The proposed architecture is construct-preserving rather than purely predictive. It emphasizes multimodal data organization, construct-level representation, temporal updating, uncertainty visibility, explainability, clinician-in-the-loop interpretation, and Responsible AI governance. The manuscript outlines candidate computational components, including multimodal fusion, temporal modeling, missing-data handling, uncertainty estimation, explainability mechanisms, and clinician-readable outputs. It also proposes operational computational representations of FIAP^® constructs and a stepwise human-supervised profiling workflow. Importantly, FIAP^®-Digital is not presented as a validated diagnostic tool, treatment recommendation system, autonomous AI model, or clinically proven platform. The constructs described remain hypothetical and require empirical validation. The article therefore provides a neuroinformatics architecture and validation roadmap, including feasibility testing, construct validity, internal and external validation, longitudinal robustness, clinician usability, fairness assessment, and uncertainty calibration. Its primary contribution is to define a responsible computational pathway for testing whether FIAP^® constructs can support future precision-stratified autism research.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1868068"},"PeriodicalIF":2.9,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13574807/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148976181","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interpretable self-supervised transformers for resting-state EEG analysis in Alzheimer's disease.","authors":"Syeda Shamaila Zareen, Nada Alzaben, Usman Ahmad, Talha Mahboob Alam, Asaad Algarni, Junsong Wang","doi":"10.3389/fninf.2026.1899573","DOIUrl":"10.3389/fninf.2026.1899573","url":null,"abstract":"<p><p>Existing EEG-based methods have been constrained by limited availability of labeled data, hand-crafted features, poor spatio-temporal modeling, sub-optimal cross-hardware performance, and lack of interpretability due to being expensive and intrusive. To address these limitations, this study introduces innovative neural signal decoding techniques for cognitive state modeling in order to enhance the potential of AI-integrated models for early detection of Alzheimer's disease. This research introduces a self-supervised spatio-temporal transformer (STT-EEG) for early detection of Alzheimer's disease from resting-state EEG. This framework includes four key components: first, self-supervised pretraining on 111 healthy controls using masked auto-encoding and temporal order prediction to learn robust generalisable representations. Second, a novel spatial attention module (SAM) that explicitly captures both long-range temporal dependencies and channel interactions, reflecting the distributed network pathology of AD; third, cross-dataset transfer learning from 64-channel BioSemi to 19-channel Nihon Kohden systems, which showed strong hardware generalization; and finally, analyses of attention rollout and channel perturbation for clinically interpretable insights. The model was trained in a subject-wise 5-fold cross-validation fashion on the SRM dataset and fine-tuned on the OpenNeuro dataset (ds004504) consisting of 36 AD and 29 CN. On the same dataset, the accuracy of STT-EEG was 96.42% for AD vs. CN classification. The most significant improvement +7.08% was made with the help of self-supervised pretraining, followed by data augmentation +6.30% and the SAM +4.86%, as was confirmed in the ablation studies. For continuous prediction of MMSE scores, the Pearson correlation of the model was 0.872 and the mean absolute error (MAE) was 2.34 points. Regions of T3-T6, P3-Pz-P4 and O1-O2 were identified as areas of attention-based interpretability, which were consistent with the known neuropathology of AD that involved the temporoparietal lobe. STT-EEG strengths include the high interpretability of the framework, its spatio-temporal attention, and the fact that STT-EEG is a self-supervised learning method and can be used in an efficient and generalizable way.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1899573"},"PeriodicalIF":2.9,"publicationDate":"2026-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13560164/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148944844","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interpretable AI for neural signal decoding in dementia: an EEG ensemble approach to differential diagnosis.","authors":"Fawad Muhammad, Irfan Ahmed Usmani, Muhammad Aamir, Mai Alduailij, Mehrez Marzougui, Rab Nawaz","doi":"10.3389/fninf.2026.1902549","DOIUrl":"10.3389/fninf.2026.1902549","url":null,"abstract":"<p><strong>Introduction: </strong>The differential diagnosis between Alzheimer's disease (AD) and frontotemporal dementia (FTD) presents a significant clinical challenge due to overlapping early-stage symptom profiles. Conventional resting-state EEG provides limited sensitivity to the impaired neural plasticity and lateralized cortical degeneration frequently observed in FTD.</p><p><strong>Methods: </strong>We developed a domain-informed heterogeneous ensemble framework incorporating dynamic neural reactivity and hemispheric asymmetry metrics from 19-channel EEG recordings acquired from 88 participants (36 AD, 23 FTD, 29 cognitively normal controls) during resting-state and photic stimulation paradigms. A neural reactivity vector (V_diff) was derived to quantify state-dependent spectral transitions. The 1,014-dimensional feature space was reduced to 200 features via recursive feature elimination, prioritizing spectral power distributions, hemispheric asymmetry indices (HAI), and stimulation-induced reactivity parameters. A weighted ensemble of Extreme Gradient Boosting (XGBoost) and Random Forest classifiers was evaluated using a subject-aware 90/10 holdout split with internal five-fold cross-validation.</p><p><strong>Results: </strong>The optimized model achieved a multi-class segment-level accuracy of 95.63% on an independently held-out test partition of 1,281 segments, with an internal five-fold cross-validation mean of 0.9846 ± 0.003. FTD-specific precision reached 0.9907. SHAP analysis identified beta-band hemispheric asymmetry and alpha-band reactivity as the principal contributors to class separation.</p><p><strong>Discussion: </strong>These findings indicate that the integration of dynamic state-transition measures with structural asymmetry proxies enhances electrophysiological discrimination between dementia subtypes. The framework provides a computationally efficient and biologically interpretable alternative to deep learning-based methodologies for EEG-driven dementia classification.</p>","PeriodicalId":12462,"journal":{"name":"Frontiers in Neuroinformatics","volume":"20 ","pages":"1902549"},"PeriodicalIF":2.9,"publicationDate":"2026-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13557975/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148929538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}