A College of Engineering team in Xin Zhang's Laboratory for Microsystems Technology trained a transformer-based foundation model on general patterns in brain MRI scans, so it can be adapted to specific diagnostic tasks with only a few labeled examples. The approach targets a common hospital constraint: plenty of imaging data but few expert-labeled cases. Radiologist Chad Farris of the Chobanian & Avedisian School of Medicine said faster reads could help flag acute conditions such as stroke. The Brink featured the work again in its September 18 research roundup.