Comparing Post-Hoc Explainable AI Methods for Interpreting Black-Box EEG Models in Depression Detection

2026/05/27 by Antonia Šarčević, Nikolina Frid · 1 voice
Computer Science · Neuroscience · Psychology · #Attribution #Deep learning #Depression (economics) #EEG and Brain-Computer Interfaces #Electroencephalography #Emotion and Mood Recognition #Feature (linguistics) #Functional Brain Connectivity Studies #Neurophysiology #cs.AI #cs.LG

paper · pdf · open access · doi:10.48550/arxiv.2605.28977

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2026/05/27 · arxiv published 2026/05/27 · arxiv updated 2026/05/27 · openalex created_date 2026/05/30 · openalex updated_date 2026/07/28

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.

Citations

• Are EEG Sequences Time Series? EEG Classification with Time Series Models and Joint Subject Training
• The Disagreement Problem in Explainable Machine Learning: A Practitioner's Perspective
• Who is afraid of black box algorithms? On the epistemological and ethical basis of trust in medical AI
• Explainable AI: A Review of Machine Learning Interpretability Methods
• A study of resting-state EEG biomarkers for depression recognition
• Self-explaining AI as an alternative to interpretable AI
• Nonlinear analysis of EEG complexity in episode and remission phase of recurrent depression
• A Benchmark for Interpretability Methods in Deep Neural Networks
• All Models are Wrong, but Many are Useful: Learning a Variable's Importance by Studying an Entire Class of Prediction Models Simultaneously
• A Unified Approach to Interpreting Model Predictions
• Learning Important Features Through Propagating Activation Differences
• Axiomatic Attribution for Deep Networks
• Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization
• Sequential rank agreement methods for comparison of ranked lists
• Visualizing and Understanding Convolutional Networks
• The Epidemiology of Depression Across Cultures
• Classifying depression patients and normal subjects using machine learning techniques and nonlinear features from EEG signal
• Affective Style and Affective Disorders: Perspectives from Affective Neuroscience
• Frontal EEG alpha asymmetry and emotion: From neural underpinnings and methodological considerations to psychopathology and social cognition

Discussions

• The optimizer dependence in neural scaling laws caught my attention. We often treat the scaling exponent as fixed, but this paper shows it varies systematically across optimizers, even with identical [bsky, 0 points, 1 comments]

Related

• Cross-Subject Depression Level Classification Using EEG Signals with a Sample Confidence Method
• Machine learning approaches in Detecting the Depression from Resting-state Electroencephalogram (EEG): A Review Study
• Representation learning for improved interpretability and classification accuracy of clinical factors from EEG
• A Systematic Review of EEG-based Machine Intelligence Algorithms for Depression Diagnosis, and Monitoring
• Machine Learning Approaches for Detecting the Depression from Resting-State Electroencephalogram (EEG): A Review Study