Microsoft AI tool diagnoses complex medical conditions step by step, outperforming general physicians in controlled tests while potentially reducing unnecessary medical investigations.
WASHINGTON: Microsoft researchers have developed a new artificial intelligence system that approaches medical diagnosis step by step, much like a human doctor, achieving promising results in identifying complex health conditions.
The Microsoft AI tool, known as MAI-DxO, demonstrated more than twice the diagnostic accuracy of general physicians using textbooks and online resources in controlled testing. The findings were published in the scientific journal Science on October 22, 2026, with further coverage emerging on October 24.
How does Microsoft’s new medical AI system work?
Unlike conventional AI models that analyse all available medical information simultaneously, the new technology follows a sequential diagnostic process.
It begins by examining a patient’s medical history, identifying relevant symptoms and determining which questions should be asked. Based on the available information, it recommends appropriate examinations and diagnostic tests before evaluating the results.
The system uses an AI orchestrator that coordinates simulated medical experts, allowing different diagnostic possibilities to be considered and refined as new information becomes available.
This approach closely resembles how doctors investigate complex medical conditions during consultations.
How accurate is Microsoft’s AI compared with doctors?
Researchers evaluated the technology using 304 challenging medical cases drawn from The New England Journal of Medicine.
According to Microsoft, the system achieved approximately 83% diagnostic accuracy, compared with 36% for general physicians using reference materials.
The Microsoft AI tool also performed strongly in simulated emergency department cases, achieving around 80% accuracy compared with 47% for physicians.
Researchers found that the system could reach diagnoses with fewer investigations and lower estimated costs.
However, these results came from controlled evaluations rather than real-world clinical practice. The comparisons also did not establish superiority over specialist medical teams.
Could AI reduce medical testing costs?
The research suggests that sequential diagnostic technology could help doctors identify necessary investigations more efficiently.
By assessing each result before recommending another test, the system may reduce unnecessary procedures while maintaining diagnostic accuracy.
This could eventually improve hospital efficiency and help healthcare professionals make better-informed decisions.
Nevertheless, clinical validation remains essential before such technology can be trusted in everyday patient care.
When will Microsoft’s AI diagnosis system be used in hospitals?
Microsoft plans to collaborate with Beth Israel Deaconess Medical Center in Boston to evaluate the technology for emergency department patient triage.
The next stage will assess whether its performance in research environments translates into practical healthcare improvements.
The Microsoft AI tool is not currently a replacement for qualified medical professionals, and its wider clinical use will depend on further testing and appropriate regulatory approvals.
The research represents an important step towards AI-assisted diagnosis, with the potential to support doctors rather than replace them.
Impact to expect
The development could accelerate the adoption of artificial intelligence in healthcare, including across the UAE and Middle East. If successfully tested in hospitals, similar systems may support earlier diagnosis, reduce unnecessary medical costs and improve access to specialist knowledge. Human medical oversight, patient safety and regulatory approval will remain essential.
