Microsoft researchers have developed an artificial intelligence system designed to approach medical diagnosis more like a doctor. According to Reuters, the system was more than twice as accurate as general physicians at working through complex medical cases in a study published in the journal Science. Unlike conventional medical AI systems that process symptoms, laboratory results and imaging together, this approach works through a sequence of decisions.
The system uses an orchestrator, a component that coordinates the diagnostic process and simulates collaboration among clinical experts. It decides which questions to ask about a patient's medical history, identifies relevant examination findings and determines which tests should be ordered. It then considers the results, requests further tests when appropriate and updates its working hypotheses. Researchers tested the system on more than 300 unusual and difficult medical cases. It also outperformed human physicians on scenarios drawn from a major database of emergency room admissions. Reuters reported that the researchers found diagnostic work-ups designed by the orchestrator cost less than those performed by human doctors.
The findings matter because medical diagnosis requires gathering evidence, deciding what to investigate and revising conclusions as new information becomes available. An AI system that can follow this process may offer a different approach from tools that mainly answer fixed medical questions. However, the reported results concern study evaluations and do not establish that the system is ready to replace doctors in routine patient care.
Microsoft researchers are continuing to develop the technology. The study's reported performance provides a basis for further evaluation of how sequential AI reasoning could support clinical decision-making.