Title: Brain Connectivity Biomarkers for PTSD: A Machine Learning Approach
Abstract: Post-traumatic stress disorder (PTSD) is a common and disabling psychiatric illness characterized by substantial clinical heterogeneity, posing challenges for diagnosis, prognosis, and treatment selection. Although PTSD symptoms are traditionally grouped into four domains (intrusion, avoidance, negative alterations in cognition and mood, and arousal/reactivity), the neural mechanisms underlying this heterogeneity remain incompletely understood. In this study, we investigated whether patterns of resting-state functional brain connectivity could predict overall PTSD severity and individual symptom domains using a machine learning framework. Resting-state functional MRI was acquired from 50 individuals with PTSD, and symptom severity was assessed using the PTSD Checklist for DSM-5 (PCL-5). Functional connectivity among 100 cortical and subcortical regions spanning the default mode, salience, executive control, and affective networks was analyzed. Principal component analysis was used for dimensionality reduction, followed by least-angle regression to predict total PCL-5 scores and domain-specific symptom severity. The model explained 29% of the variance in overall PTSD severity and significantly outperformed chance in predicting total symptom burden (p = 0.030). Prediction performance was strongest for intrusion (R² = 0.33, p = 0.002) and avoidance symptoms (R² = 0.23, p = 0.034), whereas cognition/mood and arousal/reactivity symptoms were not predicted above chance. These findings suggest that distributed patterns of functional brain connectivity capture meaningful variation in specific PTSD symptom dimensions. Although replication in larger cohorts is needed, this work highlights the promise of computational neuroimaging and machine learning for developing objective, brain-based biomarkers that may ultimately support more precise diagnosis, prognosis, and personalized treatment in PTSD.
Bio: Amin Zand Vakili, MD, Ph.D., is an Assistant Professor of Psychiatry
and Human Behavior at Brown University and a Staff Psychiatrist at the
Providence VA Medical Center. He earned his MD from Tehran University,
completed a PhD in Neuroscience at Albert Einstein College of
Medicine, and finished his psychiatry residency at Brown University
through the NIH-funded R25 Research Training Program.
Dr. Zand Vakili’s research focuses on computational psychiatry,
neuromodulation, and the development of objective biomarkers for
mental illness. His work integrates electroencephalography (EEG),
multimodal behavioral data, machine learning, and artificial
intelligence to understand treatment response and personalize care for
depression, PTSD, and other psychiatric disorders. He directs the
Recording, Decoding, and Computational Neuroscience Core at the VA
Center for Neurorestoration and Neurotechnology, where his team
develops data-driven approaches to improve diagnosis, monitor
treatment, and advance precision psychiatry.