International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

Multimodal Deep Learning Framework for PTSD Detection during Sleep via EEG and Biosignal Fusion

Authors :

Danial Eskandari Faruji¹ Amir Akhavan Saffar² Mobina Ansari Astaneh³

1- Computer engineering department, Hakim Sabzevari University 2- Faculty of Computer Engineering and Information Technology of Sadjad University 3- Faculty of Computer Engineering and Information Technology of Sadjad University

Keywords :

PTSD Detection،EEG signals،Multimodal Fusion،Sleep Analysis،CNN-BiLSTM

Abstract :

Post-Traumatic Stress Disorder (PTSD) disrupts sleep, causing nightmares, insomnia, and autonomic dysregulation that demand objective diagnostic tools. Existing EEG-based studies often focus on resting-state or unimodal analysis, missing the potential of multimodal sleep data to capture dynamic neurophysiological patterns. We introduce a novel deep learning framework that integrates EEG and ECG signals for automated, non-invasive PTSD detection during sleep, thereby enhancing clinical utility. Using a dataset of 28 subjects from Amsterdam UMC [1], we pre-process EEG data (Fpz-Cz, Pz-Oz, 100 Hz) into spectrograms via Short-Time Fourier Transform. We also extract Heart Rate Variability (HRV) from ECG signals. The core of our system is a hybrid CNN-BiLSTM architecture, which utilizes convolutional layers for spatial EEG feature extraction, bidirectional LSTMs for temporal modeling, and a multi-head attention mechanism for multimodal fusion. To address class imbalance, we use SMOTE and evaluate performance with stratified 10-fold cross-validation. The model achieves a 96% accuracy, 96% F1-score, and a 1.00 AUC, which significantly surpasses unimodal baselines (EEG-only CNN: 76.8%, SVM: 74.1%) [2]. Our ablation studies confirm that multimodal fusion boosts the detection of PTSD biomarkers, such as REM disruptions and elevated theta power. This framework provides a scalable and interpretable solution for PTSD detection, suitable for clinical diagnostics and integration into wearable devices, which advances real-time mental health monitoring.