A New Privacy-First AI Predicts COVID Severity Using X-Rays and Medical Records

Researchers built EXAM, a global AI model that predicts COVID-19 patients’ oxygen needs within 24–72 hours using chest X-rays and medical records. Instead of pooling sensitive hospital data, the team used federated learning, letting 20 hospitals train the model locally while sharing only model updates. The approach improved accuracy by 16% and generalizability by 38%, showing how privacy-preserving AI can scale healthcare collaboration.