AI-Driven Multimodal Preventive System for Adolescent Mental Health: A Randomized Controlled Trial

Abstract

Introduction

To develop and validate a multimodal AI system for early detection and prevention of adolescent mental health disorders (e.g., anxiety, depression), targeting suicide risk reduction and emotional resilience enhancement through personalized interventions.

Method

A 12-month randomized controlled trial (RCT) enrolled 1,200 adolescents (aged 12–18) to evaluate an AI platform integrating neuroimaging, behavioral sensors, and natural language processing (NLP). The AI model, trained on 10,000 multimodal datasets, assessed vocal tone, facial microexpressions, and social media text to predict suicide/depression risks (accuracy=89%). Intervention modules included AI-guided cognitive behavioral therapy (CBT), mindfulness exercises, and peer support matching. Outcomes included suicidal ideation rates, emotional resilience scores, and school absenteeism.

Results

The AI system reduced suicidal ideation by 34% (p<0.01) and improved emotional resilience by 27% (p<0.001) versus controls. Longitudinal data showed sustained benefits at 6-month follow-up, with 38% lower school absenteeism (OR=0.62, 95% CI: 0.51–0.75). Key predictors of risk included vocal tremor frequency (AUC=0.85) and repetitive negative social media posts (p=0.002). Federated learning ensured data privacy compliance during interventions.

Discussion

This AI-powered preventive system demonstrates scalable potential for early intervention in adolescent mental health, addressing digital-age psychosocial risks. By integrating real-time risk detection with targeted behavioral support, it offers a public health framework to mitigate suicide and emotional disorders, aligning with global priorities for youth mental well-being.