arXiv:2605.28977v1 Announce Type: new 
Abstract: Recent advances in deep learning have enabled increasingly accurate electroencephalography (EEG)-based classification of Major Depressive Disorder (MDD), but the decision-making processes of high-capacity models remain difficult to interpret. This study investigates multiple post-hoc explainability methods applied to an InceptionTime architecture trained for EEG-based MDD detection. The analysis includes Shapley-based, gradient-based, and perturbation-based attribution approaches: DeepSHAP, Integrated Gradients, GradCAM, Occlusion, and Permutation Feature Importance. Explainability analysis was performed within a subject-level stratified 5-fold cross-validation framework using global attribution aggregation across EEG segments and subjects. The evaluated methods revealed partially convergent attribution patterns, with recurring emphasis on frontal, temporal, and posterior EEG regions, particularly in the right hemisphere. Quantitative comparison demonstrated substantial agreement between gradient- and perturbation-based approaches, while DeepSHAP produced comparatively distinct attribution distributions. At the same time, variability between explainability methods highlighted the influence of methodological assumptions on the resulting explanations. Overall, the results suggest that different post-hoc explainability approaches capture partially overlapping relevance structures in EEG-based deep learning models for depression detection. Although the observed attribution patterns are broadly consistent with several previous EEG studies of MDD, the analysis should be interpreted as exploratory rather than evidence of definitive neurophysiological biomarkers or clinical applicability. The study highlights both the usefulness and limitations of post-hoc explainability for interpreting black-box EEG classifiers in psychiatric applications.

أظهرت الأبحاث الأخيرة استكشاف عدة طرق لشرح الذكاء الاصطناعي بعد الحدث لتفسير نماذج EEG ذات الصندوق الأسود المستخدمة في الكشف عن الاضطراب الاكتئابي الكبير (MDD). قامت الدراسة بتقييم طرق مثل DeepSHAP وIntegrated Gradients وGradCAM ضمن إطار عمل للتحقق المتقاطع من خمسة طيات، مما كشف عن أنماط نسب متسقة عبر مناطق EEG، وخاصة في نصف الكرة الأيمن.

Investigaciones recientes han explorado varios métodos de explicabilidad post-hoc de IA para interpretar modelos EEG de caja negra utilizados en la detección del trastorno depresivo mayor (TDM). El estudio evaluó métodos como DeepSHAP, Integrated Gradients y GradCAM dentro de un marco de validación cruzada estratificada a cinco pliegues, revelando patrones de atribución consistentes en las regiones EEG, particularmente en el hemisferio derecho.

Des recherches récentes ont exploré diverses méthodes d'explicabilité post-hoc de l'IA pour interpréter des modèles EEG en boîte noire utilisés pour détecter le trouble dépressif majeur (TDM). L'étude a évalué des méthodes telles que DeepSHAP, Integrated Gradients et GradCAM dans un cadre de validation croisée stratifié à cinq plis, révélant des motifs d'attribution cohérents dans les régions EEG, en particulier dans l'hémisphère droit.

Recent research has explored various post-hoc explainable AI methods to interpret black-box EEG models used for detecting Major Depressive Disorder (MDD). The study evaluated methods such as DeepSHAP, Integrated Gradients, and GradCAM within a subject-level stratified 5-fold cross-validation framework, revealing consistent attribution patterns across EEG regions, particularly in the right hemisphere.