< Back to all clusters
[TECHNOLOGY] · United States · 36 sources

started · updated

Microsoft AI system MAI-DxO outperforms doctors in complex diagnoses

Microsoft researchers have developed a medical AI system called MAI-DxO (Microsoft AI Diagnostic Orchestrator) that simulates a clinical expert panel to diagnose patients sequentially. Unlike traditional AI models that analyze all data simultaneously, this system uses an orchestrator to ask patient history questions, order specific tests, and update hypotheses as new evidence emerges.

A study published in the journal Science reported that MAI-DxO achieved an 83.2% accuracy rate in diagnosing 304 complex medical cases from the New England Journal of Medicine. In the same simulation, general physicians using textbooks and the internet achieved only 36.1% accuracy. The system also demonstrated high performance in emergency department scenarios, reaching approximately 80.3% accuracy compared to 46.8% for human physicians.

While the results are significant, researchers noted the study was conducted in a simulated environment using text-based data and has not yet been validated in real-world clinical settings. Microsoft plans to conduct real-world validation studies in partnership with Beth Israel Deaconess Medical Center. Mustafa Suleyman, head of Microsoft AI, described the development as a step toward “medical superintelligence.”

Entities

Beth Israel Deaconess Medical Center · Dominic King · MAI-DxO · Mayo Clinic · Microsoft · Mustafa Suleyman · OpenAI · Science

Claims

What the coverage asserts, and how many sources carry each claim.

Sources