Sign in Subscribe
Physicians

Microsoft Develops AI Tool That Outperforms Human Doctors at Complex Health Diagnoses

Strategic Insights

— 1 min read

The Microsoft AI team, led by British tech pioneer Mustafa Suleyman, has developed a medical AI tool called Microsoft AI Diagnostic Orchestrator (MAI-DxO) that performs better than human doctors at complex health diagnoses. The tool, which imitates a panel of expert physicians, correctly diagnosed up to 85% of cases from the New England Journal of Medicine, a rate more than four times higher than a group of experienced physicians. MAI-DxO also gets to the correct diagnosis more cost-effectively than physicians, making it a cheaper option for healthcare providers.

Key Takeaways:

  • Microsoft's AI Diagnostic Orchestrator (MAI-DxO) has been developed by the company's AI health unit, led by Mustafa Suleyman, and has shown a success rate of up to 85% in diagnosing complex health cases.
  • MAI-DxO outperforms human doctors, with a rate of over four times higher than a group of experienced physicians.
  • The AI tool gets to the correct diagnosis more cost-effectively than physicians, making it a cheaper option for healthcare providers.
  • Microsoft's approach uses a sequential diagnosis benchmark, which relies on multiple agents working together in a chain-of-debate style.
  • The company has used 304 case studies sourced from the New England Journal of Medicine to develop and test the AI tool.
  • MAI-DxO is not yet ready for clinical use, with further testing needed to assess its performance on more common symptoms.
  • Microsoft's work on MAI-DxO has the potential to reshape healthcare, empowering patients to self-manage routine aspects of care and equipping clinicians with advanced decision support for complex cases.

Statistics:

  • Up to 85% of complex health cases can be correctly diagnosed by MAI-DxO.
  • MAI-DxO outperforms human doctors, who have a success rate of around 20% in diagnosing complex health cases.
  • The AI tool gets to the correct diagnosis more cost-effectively than physicians.
  • 304 case studies sourced from the New England Journal of Medicine were used to develop and test MAI-DxO.
  • MAI-DxO uses a sequential diagnosis benchmark, which relies on multiple agents working together in a chain-of-debate style.
  • The company plans to further test MAI-DxO to assess its performance on more common symptoms.

Sources:

  • Wired: "Microsoft says its new AI tool diagnosed 85% of cases as well as doctors"
  • New England Journal of Medicine (sourced case studies)
  • Microsoft blog post (exact wording of the article)

Read more