Microsoft AI Diagnostic Tool Outperforms Physicians in Study Published in Science

Microsoft researchers have developed an AI tool that diagnoses medical conditions step by step and identified complex cases with more than twice the accuracy of general physicians, according to a study published Thursday in the journal Science. Unlike conventional medical AI models that analyze all symptoms, lab results and imaging data at once, Microsoft's system uses an "orchestrator" that first decides which patient history to gather and which physical findings matter. It then orders tests, evaluates results and requests further examinations, continuously updating its diagnostic hypotheses and care plan. The model processed more than 300 rare, difficult-to-diagnose cases and outperformed physicians relying on textbooks and internet resources. It also beat human doctors on scenarios drawn from a large emergency department database. Researchers found the orchestrator planned diagnostic workflows at lower cost than physician-ordered testing. Microsoft AI health corporate vice president Dominic King said real patients never arrive with a neatly organized case summary. The team will next partner with Boston's Beth Israel Deaconess Medical Center to test triage in emergency waiting rooms.