Code for Schizophrenia Classification from Multichannel EEG Using EEGNet with SHAP and Grad-CAM

Abstract

This repository contains the code developed for EEG-based schizophrenia classification using the publicly available RepOD resting-state EEG dataset. The implementation includes EEG preprocessing, STFT-based input generation for the 3D-CNN, EEGNet classification, participant-wise Leave-One-Subject-Out (LOSO) and Stratified Group 5-Fold validation, bootstrap confidence interval estimation, ICA-based artifact-sensitivity analysis, and explainable AI analysis using SHAP and Grad-CAM. The code accompanies the manuscript entitled “Schizophrenia classification from multichannel EEG using EEGNet with SHAP and Grad-CAM.” The RepOD EEG dataset is not redistributed in this repository and should be obtained from its original public source.