PsyDraw: A Multi-Agent Multimodal System for Mental Health Screening in Left-Behind Children
Yiqun Zhang, Xiaocui Yang, Xiaobai Li, Siyuan Yu, Yi Luan, Shi Feng, Daling Wang, Yifei Zhang
TL;DR
PsyDraw addresses the shortage of mental health professionals for left-behind children in rural China by introducing a two-stage, multi-agent system that leverages multimodal reasoning to interpret House-Tree-Person drawings. The framework decomposes the task into feature extraction and psychological reporting, producing a structured analysis $R=\psi(\phi(I))$ and a risk flag $C=\tau(R)$ through collaboration among analysis, expert, and manager agents. In a pilot with 290 drawings, PsyDraw achieves $71.03\%$ High Consistency with professionals, flags $31.03\%$ of cases as Warning, and demonstrates practical utility for early screening in resource-limited settings, while adhering to ethical safeguards and open-source dissemination (excluding prompt templates). The work highlights a scalable, safety-conscious approach to augment rural mental health screening, with limitations including cultural scope, potential biases, and the need for ongoing validation. The study contributes a concrete, two-stage, agent-based architecture for visual-linguistic interpretation of projective tests, demonstrates feasible deployment in real schools, and advances the use of AI-assisted screening tools to bridge gaps in mental health resources while emphasizing professional oversight. The practical impact lies in enabling broader reach, faster preliminary screening, and more efficient allocation of scarce professionals, with careful attention to privacy, ethics, and the role of clinicians in final diagnoses.
Abstract
Left-behind children (LBCs), numbering over 66 million in China, face severe mental health challenges due to parental migration for work. Early screening and identification of at-risk LBCs is crucial, yet challenging due to the severe shortage of mental health professionals, especially in rural areas. While the House-Tree-Person (HTP) test shows higher child participation rates, its requirement for expert interpretation limits its application in resource-scarce regions. To address this challenge, we propose PsyDraw, a multi-agent system based on Multimodal Large Language Models that assists mental health professionals in analyzing HTP drawings. The system employs specialized agents for feature extraction and psychological interpretation, operating in two stages: comprehensive feature analysis and professional report generation. Evaluation of HTP drawings from 290 primary school students reveals that 71.03% of the analyzes achieved High Consistency with professional evaluations, 26.21% Moderate Consistency and only 2.41% Low Consistency. The system identified 31.03% of cases requiring professional attention, demonstrating its effectiveness as a preliminary screening tool. Currently deployed in pilot schools, \method shows promise in supporting mental health professionals, particularly in resource-limited areas, while maintaining high professional standards in psychological assessment.