Combining Deep Learning and Graph-Theoretic Brain Features to Detect Posttraumatic Stress Disorder at the Individual...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Psychiatry

Background:

• Resting-state functional MRI (rs-fMRI) studies show altered graph metrics in posttraumatic stress disorder (PTSD).
• Graph measures' potential for individual PTSD diagnosis and feature identification remains underexplored.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) model for individual PTSD diagnosis using graph-theoretic analysis of rs-fMRI data.
• To compare the DL model's diagnostic accuracy against traditional machine learning methods.
• To identify the most discriminative graph-theoretic features for PTSD.

Main Methods:

• rs-fMRI data from 91 individuals with PTSD and 126 trauma-exposed non-PTSD controls were analyzed.
• A graph-theoretic deep learning model was developed for binary classification.
• A support vector machine (SVM) classifier was used as a benchmark for comparison.

Main Results:

• The DL model achieved higher accuracy (80%) in discriminating individuals with PTSD compared to SVM (57.7%).
• Top discriminant features were located in the default mode, central executive, and salience networks.
• Nodal efficiency in the left fusiform gyrus correlated negatively with PTSD symptom severity (Clinician Administered PTSD Scale).

Conclusions:

• Deep learning based on graph features shows significant promise for aiding in the individual diagnosis of PTSD.
• Graph-theoretic analysis combined with DL can identify neural correlates of PTSD.
• This approach may enhance clinical assessment and understanding of PTSD.