Integrating Causal Inference with Digital Twin Modeling for Individualized Mental Health Research

The diagnosis and treatment of mental disorders face unprecedented dual challenges of "reproducibility crisis" and "therapeutic ceiling." Current diagnostic frameworks (e.g., DSM/ICD)and dimensional approaches (e.g., Research Domain Criteria [RDoC]) have limited biologicalvalidation and poor predictive power for individual outcomes. This theoretical frameworkreview systematically proposes and evaluates the causally integrated digital twin as a novelparadigm for mental disorders research, marking a fundamental shift from descriptive classification to individualized dynamic simulation and causal mechanism discovery. It aims tosynthesize existing evidence and provide a coherent, testable roadmap for future research.We conducted a systematic scoping review (PRISMA-ScR) of 116 studies (2021 to 2026)focusing on deep learning applications involving multimodal fusion, causal inference, ordigital twin concepts. We employed a structured theoretical synthesis method to integratefindings into a novel conceptual framework, which was refined through expert consultation.The proposed paradigm integrates three technical pillars: (1) multimodal data fusion, (2)causal inference, and (3) dynamic simulation. While preliminary evidence supports technicalfeasibility, clinical validation remains limited. Common methodological limitations includelack of standardization, small sample sizes, and untestable causal assumptions. Comparative analysis with traditional diagnostic tools reveals significant limitations in sensitivity,specificity, and dynamic prediction. The causally integrated digital twin paradigm representsa transformative theoretical framework that could address current limitations by enablingindividualized, mechanism-based modeling. This review provides a unified roadmap for futureresearch, outlining specific validation pathways and addressing key methodological, clinical,and ethical challenges that must be overcome through rigorous interdisciplinary collaboration