The rapid development of artificial intelligence has expanded the application of neural networks in the identification and intervention of mental disorders, demonstrating their potential in symptom detection, emotional state recognition, and assisted intervention. This narrative review summarizes research published between 2010 and 2025 on neural network-based approaches to mental disorder identification and intervention management. From the perspectives of model types, application scenarios, and developmental pathways, it outlines the applications of different models in structured data modeling, speech and facial image analysis, physiological signal recognition, natural language processing and social media text mining, as well as intelligent intervention and management platforms. Existing studies suggest that these approaches offer certain advantages in improving early screening and risk prediction, and in supporting the individualized and digital development of mental health services. Nevertheless, challenges remain regarding model interpretability, data privacy protection, cross-cultural adaptability, model generalizability, and ethical and regulatory considerations.