Microsoft AI system outperforms physicians on medical diagnosis tests
Microsoft AI says its diagnostic system, MAI-DxO, reached 83% accuracy on 304 complex cases from The New England Journal of Medicine, compared with 22% for physicians without reference resources and 36% with them. On typical emergency department cases from MIMIC-IV-ED, the system reached 80% accuracy versus 47% for physicians, and the company says it did so at half or one third of the cost. The work is published in Science, and Microsoft AI is open-sourcing parts of its benchmark, SDBench. The company also plans a prospective study with Beth Israel Deaconess Medical Center in Boston, where patients will interact with the system directly.