To yield clinical utility in mental disorder identification individually, we used a multiple instance learning-based method to construct a digital model based on clinical MRI scans for automated detection of patients with psychiatric disorders. An accuracy of 84% was achieved in the primary dataset with 19453 subjects, and 76% in external dataset with 600 subjects. A higher sensitivity was achieved in identifying high-risk subjects than self-scaled questionnaires (71.1% vs 22.2%) in 148 prospectively recruited college students. With a complete workflow of development and validation, the constructed model is more practical to be translated in high-risk subject screening among vulnerable populations.
This abstract and the presentation materials are available to members only; a login is required.