Research
159d agoSubtype-Guided Contrastive Learning Diagnoses Brain Disorders
The new BrainSCL framework tackles the challenge of heterogeneity in mental‑disorder data by learning latent subtypes through contrastive learning. By fusing clinical text with adaptive graph structures derived from BOLD signals, the method uncovers robust subtype prototypes that improve diagnostic accuracy worldwide.
The Signal
This approach offers a scalable, data‑driven tool for clinicians and researchers across diverse health systems.