Resting-state EEG functional connectivity predicts post-traumatic stress disorder subtypes in veterans

Area of Science:

• Neuroscience
• Psychiatry
• Computational Neuroscience

Background:

• Post-traumatic stress disorder (PTSD) is a heterogeneous condition, complicating biomarker identification.
• Previous electroencephalography (EEG) studies on PTSD have yielded inconsistent results due to limited feature analysis.
• Quantifiable biomarkers are needed for targeted PTSD treatment.

Purpose of the Study:

• To develop a machine learning framework for analyzing a wide range of EEG biomarkers in PTSD.
• To identify EEG features capable of characterizing PTSD and its potential subtypes.
• To improve the accuracy of PTSD classification using EEG data.

Main Methods:

• Recorded resting-state EEG (eyes-closed and eyes-open) from 202 combat-exposed veterans (PTSD and controls).
• Computed spectral, temporal, and connectivity EEG features.
• Employed logistic regression, random forest, and support vector machines with feature selection and cross-validation for classification.

Main Results:

• Classifiers achieved up to 62.9% balanced test accuracy for predicting PTSD.
• Identified two PTSD subtypes: one with similar EEG to controls, another with increased global functional connectivity.
• Classifying the high-connectivity PTSD subtype achieved 79.4% accuracy.
• Alpha connectivity in attention networks was crucial for prediction and correlated with arousal symptoms.

Conclusions:

• Unsupervised subtyping using a novel machine learning framework can delineate PTSD heterogeneity.
• This approach significantly improves machine learning-based PTSD prediction.
• The study highlights potential quantifiable EEG biomarkers for PTSD, particularly functional connectivity patterns.