AI-Driven Multimodal Therapeutic System with Virtual Reality Integration for Anxiety Disorders in Adolescents

Anxiety disorders are increasingly prevalent, especially in the Middle East, with a prevalence of 6.33%, where social stigma often discourages individuals from seeking medical help. Traditional clinical interviews frequently lead to misdiagnoses, and single-approach therapies typically address only one or two of the three dimensions of anxiety (cognitive, psychological, and physiological). To address these challenges, a mobile application was developed to provide accessible, effective mental health support for adolescents aged 13–19. The app features an AI chatbot that diagnoses users based on the DSM-V and SCIDRV-V criteria to design an unstructured diagnostic model integrating four therapeutic approaches: Cognitive Behavioral Therapy (CBT), Acceptance and Commitment Therapy (ACT), Mindfulness-Based Cognitive Therapy (MBCT), and Mindfulness-Based Stress Reduction (MBSR) by utilizing the flexibility of the Gemini API to address the multidimensional nature of anxiety. The app includes a journaling page to allow adolescents to document thoughts, a history page to provide context for diagnosis, and video-based group therapy to enhance social skills. To ensure a successful group therapy experience, the Adolescent Group Therapy Readiness Questionnaire (AGTRQ), an adapted form of the Group Therapy Questionnaire (GTQ) and Group Readiness Questionnaire (GRQ), is filled by users to place them in suitable groups. Additionally, the app supports all languages, including Egyptian dialects, and ensures privacy through encrypted chat histories, with psychiatrist access restricted to diagnostic purposes. Additional features include AI-driven facial-expression and voice-tone analysis for real-time emotional feedback and Virtual Reality Exposure Therapy (VRET) to enhance engagement. Tested on 587 participants (217 experimental, 370 control), the app achieved a diagnostic sensitivity of 91%, specificity of 93.5%, a positive predictive value of 89.2%, a negative predictive value of 94.8%, and an overall diagnostic accuracy of 92.7%. Its Matthews Correlation Coefficient was 84.4%, with an F1 score of 90.2%. Wilcoxon Signed-Rank Test showed that the experimental group demonstrated a clinically significant SCAS reduction (ΔM=-7.5, 95% CI [-8.2, -6.8]), exceeding the 4–6 point Minimal Clinically Important Difference (MCID) threshold used in pediatric anxiety measures, with anxiety reductions 4.5× greater than the control group, outperforming standalone iCBT. The app demonstrates its potential as a scalable, evidence-based solution for regions with stigma or limited access to mental health services, offering both accessibility and effectiveness in managing adolescents’ anxiety disorders.