Acute Stress Disorder Detection using Machine Learning based on resting-state fMRI.

Area of Science:

• Neuroimaging and Machine Learning
• Psychiatric Diagnostics

Background:

• Early diagnosis of Acute Stress Disorder (ASD) is crucial to prevent its progression to post-traumatic stress disorder (PTSD).
• Current diagnostic methods for ASD exhibit subjectivity in assessing trauma responses and stress severity.
• Resting-state functional magnetic resonance imaging (rs-fMRI) offers a potential objective biomarker for neurological conditions.

Purpose of the Study:

• To develop and validate a machine learning model for the early detection of ASD.
• To utilize rs-fMRI data and advanced feature extraction techniques for objective ASD assessment.
• To identify specific brain regions and imaging features indicative of ASD.

Main Methods:

• rs-fMRI data from 48 subjects and PTSD Check List - Civilian Version (PCL-C) scores were analyzed.
• Frequency-domain and graph-based features were extracted from blood-oxygen-level dependent (BOLD) signals across cortical and subcortical regions.
• A multi-layer perceptron model was trained and evaluated using a leave-one-subject-out cross-validation scheme.

Main Results:

• Eighteen extracted features showed significant differences (p<0.05) between groups.
• The machine learning model achieved high diagnostic performance: 91.7% accuracy, 96.8% sensitivity, and 82.4% specificity.
• The Right Accumbens and Lingual Gyrus demonstrated significant impact and large effect sizes within the predictive model.

Conclusions:

• Machine learning analysis of rs-fMRI data provides a highly accurate and objective method for detecting ASD.
• This approach can overcome the limitations of subjective clinical assessments.
• The findings highlight the potential of neuroimaging biomarkers for early ASD diagnosis and intervention.