Major psychiatric disorders (MPD), including Major Depressive Disorder (MDD), Bipolar Disorder (BD), and Schizophrenia (SCZ), are complex neuropsychiatric illnesses and leading causes of global morbidity. Current diagnoses rely on subjective behavioral evaluations and clinical symptoms, often resulting in misdiagnoses and delayed treatment. This research aims to develop a foundation model that integrates multimodal data, including neuroimaging (functional and structural MRI), clinical records, and genetic information, to provide an objective and accurate diagnosis of MPDs. The core objectives are:
1. Establish a robust foundation model that can integrate large-scale multimodal patient data to enable precision diagnosis of MPDs and address gaps in objective diagnostic criteria.
2. Identify disease subtypes and biomarkers through multimodal feature clustering, differential expression analysis, and neuroimaging techniques.
3. Uncover the pathophysiological mechanisms of MPDs by utilizing AI tools such as model interpretability and the reasoning capabilities of large language models (LLMs). This will help create a diagnostic benchmark for MPDs and a diagnostic spectrum that reflects the heterogeneity and comorbidity of psychiatric disorders.
