Machine learning for post-traumatic stress disorder identification utilizing resting-state functional magnetic...

Area of Science:

• Neuroimaging
• Machine Learning
• Psychiatry

Background:

• Early detection of post-traumatic stress disorder (PTSD) is vital for effective patient recovery.
• Understanding brain region performance deviations and functional connectivity in PTSD is essential.

Purpose of the Study:

• To investigate performance deviations in specific brain regions of individuals with PTSD compared to healthy controls.
• To assess interregional functional connectivity in PTSD using resting-state functional magnetic resonance imaging (rs-fMRI).
• To apply machine learning techniques for identifying PTSD and healthy controls.

Main Methods:

• Extracted rs-fMRI data from 10 regions of interest (ROI) in 14 PTSD subjects and 14 healthy controls.
• Utilized ANOVA for performance level assessment and Pearson's correlation for functional connectivity analysis.
• Employed machine learning algorithms including logistic regression, K-nearest neighbor (KNN), and support vector machine (SVM) with various kernels for classification.

Main Results:

• Observed significant performance deviations in brain regions of PTSD patients compared to healthy controls.
• Identified significant positive or negative functional connectivity among ROI in PTSD brains.
• KNN and SVM with radial basis function kernel achieved high classification accuracies (up to 99.2%) on training, validation, and testing datasets.

Conclusions:

• Brain region performance and functional connectivity analysis using rs-fMRI can help discriminate PTSD subjects.
• Machine learning algorithms, particularly KNN and SVM, show promise for accurate PTSD identification.
• Findings may guide future research and clinical applications for PTSD diagnosis.