Di Gao ^* Guanghao Yang Jiarun Shen Fang Wu

• China University of Mining and Technology (Beijing), Beijing, China

Adolescence is a critical developmental stage marked by significant physical, psychological, and behavioral changes, which heightens the importance of assessing and managing health risks during this period. Traditional health risk assessment methods struggle to accurately predict mental and behavioral health risks in adolescents, primarily due to the complexity of brain function data and the challenge of acquiring high-quality annotated functional magnetic resonance imaging (fMRI) data. To address these challenges, this study proposes a novel approach that integrates fMRI with deep learning techniques, specifically utilizing a multi-sequence two-dimensional convolutional autoencoder (2DCNN-AE) and multi-scale asynchronous correlation information extraction. The proposed method effectively extracts spatial and temporal features from fMRI data and reconstructs samples, thereby reducing the cost of model development. Experimental evaluations on the Adolescent Risk Behavior (AHRB) dataset, comprising 174 participants aged 17-22, demonstrated that the proposed method achieved a precision of 83.116%, recall of 84.784%, and an F1-score of 83.942%, outperforming existing methods across most evaluation metrics. These results highlight the potential of the method to accurately assess adolescent health risks, providing a robust framework for early intervention.