AI-ENABLED NEUROBIOLOGICAL DIAGNOSTIC MODELS FOR EARLY DETECTION OF PTSD AND TRAUMA DISORDERS

Abstract

This systematic literature review examines the development and application of artificial intelligence (AI)-enabled neurobiological diagnostic models for the early detection of post-traumatic stress disorder (PTSD) and trauma-related disorders. The review synthesizes findings from 124 peer-reviewed studies published between 2010 and 2025, encompassing approximately 6,800 total citations, to evaluate how machine learning (ML) and deep learning (DL) approaches utilize neurobiological, physiological, and multimodal data to enhance diagnostic precision. Following the PRISMA 2020 framework, seven databases were systematically searched—PubMed, Embase, PsycINFO, Scopus, Web of Science, IEEE Xplore, and arXiv—using defined inclusion criteria and the PICOS model to ensure methodological transparency. Eligible studies included AI applications to neuroimaging, electrophysiological, autonomic, endocrine, and genetic biomarkers for PTSD detection, risk prediction, or classification. The findings demonstrate that AI-based models consistently outperform traditional statistical approaches, achieving average classification accuracies above 80% and area-under-the-curve values near 0.85. Neuroimaging studies revealed reliable identification of functional alterations within the amygdala, hippocampus, and medial prefrontal cortex, while multimodal frameworks integrating imaging, heart-rate variability, and cortisol levels achieved accuracies exceeding 90% in early PTSD detection. Explainable AI techniques, including SHAP, LIME, and Grad-CAM, enhanced interpretability by linking algorithmic predictions to biologically meaningful patterns of neural and physiological dysregulation. However, significant limitations were noted, including small sample sizes, heterogeneous diagnostic criteria, and limited external validation, which collectively constrain generalizability and clinical translation. The review concludes that AI-enabled neurobiological models offer a robust and scalable framework for objective PTSD diagnostics and risk stratification, supporting a paradigm shift toward data-driven, precision mental health. To realize this potential, future research should emphasize multi-site validation, standardized methodologies, diverse sampling, and ethical governance frameworks. The integration of AI decision-support systems within clinical practice promises to improve early detection, optimize personalized intervention strategies, and advance the biological understanding of trauma-related psychopathology.