UCSF researchers programmed a machine-learning algorithm to diagnose early-stage Alzheimer’s disease. The algorithm used PET scans – a common type of brain scan.