ECG Signal Analysis for Detection and Diagnosis of Post-Traumatic Stress Disorder: Leveraging Deep Learning and...

Area of Science:

• Computational neuroscience
• Medical informatics
• Cardiovascular signal processing

Background:

• Post-traumatic stress disorder (PTSD) is a critical mental health condition with potential cardiovascular complications.
• Early and accurate diagnosis of PTSD is essential for effective treatment.
• Current diagnostic methods may be invasive or time-consuming.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) system for PTSD detection using electrocardiogram (ECG) signals.
• To compare the performance of deep learning (DL) models with traditional machine learning (ML) classifiers.
• To investigate the impact of ECG signal segment length on classification accuracy.

Main Methods:

• ECG signals were converted into time-frequency images using Continuous Wavelet Transform (CWT).
• Deep learning models (AlexNet, GoogLeNet, ResNet50) and traditional ML classifiers were employed.
• Performance was evaluated using various segment lengths (5s, 10s, 15s, 20s).

Main Results:

• ResNet50 achieved the highest accuracy (94.92%) with a 5-second ECG segment.
• Deep learning models significantly outperformed traditional ML approaches.
• The Area Under the Curve (AUC) for ResNet50 reached 0.99, indicating high diagnostic capability.

Conclusions:

• CNN-based models using ECG time-frequency representations show high accuracy for PTSD classification.
• Shorter ECG signal segments (5s) yield optimal performance.
• This AI-driven approach offers a promising non-invasive tool for PTSD diagnostic support.