A new artificial intelligence system developed by Microsoft achieved an 83 per cent diagnostic accuracy rate on complex medical cases during recent evaluations, according to a company statement.
By comparison, physicians using textbooks and online resources reached a 36 per cent accuracy rate on the same assessments. Doctors working without external reference materials recorded an accuracy rate of 22 per cent.
Researchers tested the technology, known as the MAI-DxO, using 304 complex cases drawn from The New England Journal of Medicine. The software employs multiple artificial intelligence agents working together to assess symptoms, select diagnostic tests and reach conclusions.
The platform also performed strongly on routine emergency department scenarios. It achieved an 80 per cent accuracy rate on typical emergency cases, while physicians managed 47 per cent.
Microsoft stated that the technology could lower diagnostic testing costs by selecting fewer or less expensive procedures while maintaining high accuracy.
In a cost-optimized configuration, the system reached an 81 per cent accuracy rate on complex cases at less than half the estimated cost of comparative systems.
The company described the findings as progress toward medical superintelligence, referring to its vision of artificial intelligence systems capable of surpassing leading clinical teams in medical reasoning and care coordination.
The tests relied on simulated cases rather than real-world clinical practice, and the physician comparison did not include specialist teams.
Microsoft acknowledged that evaluation in clinical environments is necessary to establish how the technology performs in actual patient care.
The technology firm plans to evaluate the system through a prospective study with Beth Israel Deaconess Medical Center in Boston. The study will compare its diagnostic assessments with those of physicians in hospital settings.
