Microsoft AI Solves Complex Medical Diagnoses with 83% Accuracy — More Than Twice as Accurate as Doctors

Microsoft's AI has significantly outperformed physicians in diagnostic accuracy. According to a study published Oct. 22 in the journal Science, Microsoft's AI diagnostic system, MAI-DxO, correctly diagnosed 83% of complex medical cases in a simulation test. In the same test, general practitioners — even with access to reference materials — achieved an accuracy rate of just 36%, meaning the AI scored more than twice as high.

Microsoft dubbed the technology the Medical AI Diagnostic Orchestrator (MAI-DxO). It operates like a panel of physicians, with multiple AI agents debating difficult cases. Rather than offering an answer from the start, it begins with limited information, asks follow-up questions and considers which tests are needed. Each time new information comes in, the agents challenge one another's conclusions to refine the judgment. MAI-DxO uses multiple AI agents to evaluate symptoms, recommend tests and arrive at a diagnosis.

Take, for example, a patient suffering from unexplained fatigue and abnormal test results. A doctor might settle on one possibility and order additional tests, but the system reviews multiple possibilities simultaneously and revisits its initial judgment as the evidence shifts. Microsoft said the approach is especially useful when symptoms span multiple conditions.

In the study, published Oct. 22 in Science, the researchers evaluated the technology using 304 complex medical cases drawn from the New England Journal of Medicine. However, the study was conducted in a simulated environment, and the physicians involved were not a group of specialists jointly reviewing difficult cases.