Individual-specific functional connectivity predicts clinical symptoms severity in patients with post-traumatic stress disorder.

backgroundDespite significant group-level findings on functional connectivity (FC) alterations in patients with Post-Traumatic Stress Disorder (PTSD), previous studies have failed to establish reliable neuroimaging biomarkers for diagnosis or symptom prediction. This exploratory study aims to investigate the potential predictive value of resting-state brain network FC at both individual-specific and group levels for clinical symptoms in PTSD patients, and to identify predictive FC biomarkers in specific functional networks.

methods45 PTSD patients diagnosed according to Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) criteria (12 males/33 females, aged 28.64 ± 6.87 years) were enrolled. All participants underwent high-resolution T1WI and resting-state functional MRI sequence acquisitions using a 5.0 T MRI system. FC matrices were generated based on both individual-specific parcellation and group-level atlases. A support vector machine for regression model combined with leave-one-out cross-validation was applied to evaluate the predictive performance of FC for the Posttraumatic Stress Disorder Checklist-5 (PCL-5) scores in PTSD patients. Key functional connections identified as predictors of symptom severity were analyzed, and each individual-specific subregion was classified based on its membership within one of seven major brain networks. The large-scale brain network properties significant for predicting PTSD symptoms were further explored.

resultsFunctional connectivity derived from individual-specific parcellation significantly predicted PCL-5 scores in PTSD patients (r = 0.5275, p = 0.0010), whereas the association at the group-level was significantly reduced (r = -0.0360, p = 0.3710). Further analysis revealed that the key predictive functional connections primarily involved between-network connectivity and with dorsal attention network (DAN) contributed most significantly to the prediction of symptom severity. Interestingly, FC between networks was overall positively correlated with higher PCL scores (r = 0.5265, p = 0.0002).

conclusionsIndividual-specific FC analysis demonstrated preliminary evidence of higher accuracy and relevance in predicting clinical symptoms of PTSD, with DAN contributing most significantly to the prediction. The predictive effect was primarily linked to bidirectional alterations in between-network connectivity, with a net positive correlation observed on average. These findings provide evidence for the utility of using individualized brain region homologies in predicting symptom severity within PTSD. CLINICAL TRIAL NUMBER: Not applicable.