The GAI Technology for Healthcare Analytics in Factors Influencing Exercise Addiction

This study aimed to identify neurobiological biomarkers for the early detection of exercise addiction (EA) by integrating generative artificial intelligence (GenAI) with healthcare analytics. Specifically, it sought to overcome limitations of conventional approaches in early risk identification and personalized intervention planning. A Transformer-based multimodal fusion architecture was developed to integrate physiological, behavioral, and psychological data, while cross-modal attention mechanisms facilitated dynamic interactions among heterogeneous information sources. In addition, a deep learning–enhanced magnetic resonance imaging (MRI) preprocessing framework was designed to combine conventional neuroimaging techniques with advanced representation learning methods, transforming raw MRI data into high-quality brain network representations. A total of 102 participants were recruited for experimental validation. High-resolution T1-weighted anatomical images were acquired using a Siemens 3T MRI scanner, and model performance was evaluated through 10-fold cross-validation. The proposed framework achieved an accuracy of 86.7%, sensitivity of 84.3%, specificity of 88.9%, and an F1-score of 85.5% for EA classification. Notably, it demonstrated superior performance in identifying moderate-risk cases, achieving a sensitivity of 89.6%. These findings indicate that integrating GenAI-driven design principles with multimodal healthcare data provides an effective strategy for the early detection and risk stratification of EA. The proposed physiological–psychological dual-support framework offers practical guidance for implementation and may contribute to the promotion of sustainable exercise behaviors. Nevertheless, the framework is currently limited to early risk identification and stratification. While the framework provides biomarker-based evidence to inform subsequent personalized interventions, it does not constitute a fully deployable clinical intervention system.