Microsoft’s AI Diagnosis Tool Surpasses General Physicians in Complex Cases

Microsoft researchers reported in a study published Thursday in Science that an artificial intelligence tool which mimics a clinician’s step‑by‑step diagnostic process was more than twice as accurate as generalist physicians when tackling complex medical cases.

The system employs an “orchestrator” that decides which patient questions to ask, which exams and tests to order, and how to integrate findings from a simulated panel of clinical experts. This differs from many existing AI models that attempt to generate a diagnosis from all patient data at once.

Microsoft’s research showed the AI orchestrator was more effective in diagnosing complex cases at a lower cost. (NVIDIA)

When presented with realistic emergency‑room scenarios drawn from a large clinical database, the AI outperformed human doctors. The model diagnosed more than 300 highly unusual and confusing medical dilemmas, achieving over double the accuracy of general physicians who relied on textbooks and the internet.

“AI has become very good at getting a 100% pass rate on those,” said Dr. Dominic King, corporate vice president of Microsoft AI’s health unit, which is developing the technology. “But of course, that’s not real‑world. No patient ever came to see me with a nice summary.”

The researchers also measured efficiency. “We’re not just measuring the accuracy of the systems. We’re also measuring how judiciously and carefully it goes through a diagnostic work‑up,” said Harsha Nori, head of AI at Microsoft AI Health.

Future validation will occur in partnership with Beth Israel Deaconess Medical Center in Boston, where the tool will be evaluated for triaging patients in emergency‑department waiting rooms.

Microsoft will next test the AI model and orchestrator in an emergency room setting. (Philips)

Reuters contributed to this report.

Source link