Multivariate Machine Learning Analysis of M-ECG-derived Heart Rate Variability in TBI Veterans, With and Without Current PTSD and Additional Psychiatric Comorbidity

Traumatic brain injury (TBI) and posttraumatic stress disorder (PTSD) frequently co-occur in Veterans, producing overlapping symptoms and shared autonomic dysregulation. Heart rate variability (HRV) offers a noninvasive measure of autonomic function. This study explored whether multivariate HRV features extracted from MEG-derived electrocardiogram (M-ECG) signals contained information related to group classification of Veterans with TBI alone (TBI-alone; n = 42) versus those with current PTSD, with or without additional psychiatric comorbidity (TBI+PTSD; n = 40). Time-domain, frequency-domain, geometric, and nonlinear HRV metrics were analyzed using a nested cross-validated Random Forest classifier, with Boruta-based feature selection and SHapley Additive exPlanations for model interpretability. The classifier demonstrated modest exploratory discrimination (Random Forest AUC = 0.663, 95% CI: 0.540-0.778). LF/HF ratio, low-frequency proportion, and approximate entropy were among the features contributing most to distinguishing the TBI+PTSD group, suggesting that multivariate HRV analysis of M-ECG signals may capture subtle, yet potentially informative, autonomic patterns associated with current PTSD and psychiatric burden among Veterans with TBI. The modest classification performance indicates that these findings should be considered preliminary and hypothesis-generating. Larger, independent studies are needed to determine the robustness and potential clinical relevance of these findings.