Mental and neurological disorders has been raising a major health concern, impacting more than 3.4 billion people globally and are the primary cause of illness disability and even early mortality (Neurology, 2024), in which 443 million ones recognized to be reduced health due to them in 2021 (Neurology, 2024). Anxiety and depression are two mental health conditions that affect around 970 million people worldwide and significantly increase the global disability rate. In spite of the fact that only less than 3 % of national health expenditures worldwide have been spent for mental health services the economic consequences are significant, with productivity losses from mental health issues surpassing direct healthcare costs (WHO, 2022).

Treatment of mental health is positively changing thanks to the integration of data science and artificial intelligence (AI) in psychiatry. AI systems, especially machine learning-based ones, are have been used to evaluate complicated datasets, such as genetic and neuroimaging data, to improve diagnostic accuracy. These tools have the ability to detect patterns that conventional techniques might miss, resulting in more precise and earlier diagnoses. They can also design individualized treatment programs which are appropriate to the particular needs of each patient (Whiteford et al., 2015). Moreover, AI models can comprehensively access figures for past treatment responses and automated assessment tools, thereby enabling fast modifications based on real-time data, rather than using limited methods used in psychiatric care like trial-and-error. Therefore, they can predict the most optimal interventions as well as improving and tracking their effectiveness (Whiteford et al., 2015).

With the progression of science and technology, many studies on the same topic are carried out by different research groups around the world. This work contributes to improving the accuracy and reproducibility of a research result. However, having many studies related to the same topic also causes certain difficulties for scientists. The existing studies are relatively fragment, related to different disorders, datasets and evaluation metrics, without a comprehensive analysis to come up with a set of criteria to select suitable AI tools, data processing methods as well as performance metrics for specific mental health conditions. Beside, many the AI-related studies relying on small, imbalanced, or geographically limited datasets also raise concerns about generalizability and reproducibility. Therefore, it is important to conduct systematic review studies, especially before starting to design a new research. The results of the systematic review will provide the most up-to-date and comprehensive view of current issues in the research direction, contributing to assist scientists in identifying research gaps and make decisions about choosing research direction, ultimately saving resources and time, as well as enhancing the quality of the research to be carried out. Specifically, in this study, our main purpose was to synthesize published research results of the application of AI techniques in data analysis to diagnose mental health disorder, thereby providing suggestions on suitable AI tools and methods for diagnosing mental health disorder, to guide clinical trials, and bring positive results into practical applications. In addition, we also found challenges in the use of data and the results of applying AI techniques, accordingly suggesting potential directions for future research on this issue.