earticle

영어

Post-Traumatic Stress Disorder (PTSD) is a serious psychiatric condition that requires objective and accurate diagnosis for effective early intervention. In this study, we propose a deep learning-based computer-aided diagnosis (CAD) system using event-related potentials (ERP) to distinguish individuals with PTSD from healthy controls. We introduce a novel model, TIME-CNN (Temporal Integration with Multi-scale dEcoding CNN), designed to extract temporal features through multi-scale depthwise convolutions and residual connections. EEG data were collected during an auditory oddball task from 51 PTSD patients and 39 matched healthy controls. As a result, the proposed TIME-CNN outperformed its shallower counterpart (Shallow TIME-CNN) in both classification accuracy (86.05% vs. 76.40%) and training time (7.5 vs. 12.5 hours). These findings demonstrate the effectiveness and practicality of the TIME-CNN model for ERP-based PTSD diagnosis.