Comparing Post-Hoc Explainable AI Methods for Interpreting Black-Box EEG Models in Depression Detection
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 Permutatio
The increasing sophistication of black-box AI models in critical applications like healthcare necessitates concurrent advancements in explainability to ensure trust and adoption.
Understanding how AI models diagnose conditions like depression from complex data such as EEG is crucial for clinical acceptance, regulatory approval, and responsible AI deployment in medicine.
The ability to interpret and validate AI decisions in healthcare applications will accelerate their integration into clinical practice and potentially improve diagnostic accuracy and treatment selection.
- · AI explainability researchers
- · Healthcare AI developers
- · Patients with MDD
- · Neurology and psychiatry clinicians
- · Black-box AI models without explainability
- · Traditional diagnostic methods lacking objective biomarkers
Improved trust and adoption of AI-powered diagnostic tools in mental health.
Faster and more accurate diagnosis of conditions like MDD, leading to earlier intervention.
Personalized treatment plans for neurological and psychiatric disorders based on explainable AI insights, potentially reducing trial-and-error approaches.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG