Microsoft’s AI diagnostic tool MAI-DxO achieved 83% accuracy in complex medical cases during tests, outperforming doctors who scored 36% in similar simulations, according to multiple sources.
Microsoft’s artificial intelligence diagnostic system, MAI-DxO, demonstrated 83% accuracy in identifying complex medical conditions during controlled tests, surpassing the 36% success rate of doctors using medical references, according to reports from multiple outlets.
How the AI Diagnoses Medical Cases
The system mimics a team of physicians by sequentially gathering patient history, selecting appropriate tests, and updating diagnostic hypotheses based on emerging data. This step-by-step approach contrasts with traditional AI models that analyze all symptoms at once, as explained by sources including Al Jazeera and Roaya News.
The AI system, developed by Microsoft AI, employs an “orchestrator” to coordinate multiple AI agents that simulate the collaborative decision-making of a medical team. This includes asking targeted questions, recommending tests, and revising hypotheses as new data emerges. According to Veto Gate, the system was tested on 304 complex cases drawn from The New England Journal of Medicine and emergency department databases, with results published in the journal Science on October 22, 2026.
Performance Metrics and Cost Analysis
In tests involving 304 complex cases, MAI-DxO achieved 83% accuracy compared to 36% for doctors with reference materials, while emergency cases saw 80% accuracy versus 47% for physicians, according to Al Masry Al Youm and Veto Gate.
However, the AI’s performance varied with underlying models. Using GPT-5.5, accuracy rose to 87% in complex cases, but emergency room results remained around 85%. Experts caution that real-world implementation requires further validation, as noted by Al Jazeera. The system’s cost efficiency was also context-dependent: a configuration targeting cost reduction achieved 81% accuracy at $2,241 per case, compared to $4,725 for physician-led pathways, per Microsoft’s internal analysis cited by Youm7.
Expert Reactions and Future Development
Dominic King noted, The AI excels in structured scenarios, but clinical practice involves unstructured, dynamic variables. Harsha Nuri, head of AI at Microsoft Health, added, We measure not just accuracy but the rigor of the diagnostic process.
The collaboration aims to evaluate the AI’s performance in real-world workflows, including patient interactions and physician comparisons.
While some sources highlight the AI’s efficiency, Al Jazeera notes that the AI’s real-world effectiveness remains unproven, requiring extensive clinical trials before widespread adoption. The company's ongoing research highlights the balance between innovation and ensuring safety, as emphasized in Al Jazeera.
>