Key facts
- Microsoft's AI diagnostic system, MAI-DxO, achieved 83% accuracy on complex medical cases in simulation.
- Generalist physicians achieved 36% accuracy on the same complex medical cases.
- The AI tool uses an "orchestrator" to coordinate AI agents in a step-by-step diagnostic process.
- Microsoft is partnering with Mayo Clinic to train a healthcare foundation model.
- The tool will be tested in real-world hospital settings with Beth Israel Deaconess Medical Center.
Microsoft researchers have developed an AI diagnostic system, MAI-DxO, that demonstrated superior performance in simulations, correctly diagnosing complex medical cases 83% of the time compared to 36% for generalist physicians. The system utilizes an "orchestrator" to coordinate AI agents, mimicking a doctor's step-by-step approach to questioning patients, ordering tests, and updating hypotheses. This method significantly improved the accuracy of models like OpenAI's GPT-5.5, according to a study published in Science. The researchers also noted that the AI-driven diagnostic process was less costly than human doctors using traditional resources.