Microsoft researchers on Monday unveiled a new artificial intelligence (AI) system that could diagnose more accurate patients than human doctors. Microsoft AI diagnostic orchestrator (MAI-DXO) was dubbed, including several AI models and a structure that allows it to undergo patient symptoms and history to suggest relevant tests. Depending on the results, it suggests a possible diagnosis. The Redmond-based tech veteran highlighted that apart from the accuracy of the diagnosis, the system is also trained to be cost effective in terms of tests.
Microsoft develops benchmark to test the performance of Mai-DXO
One in Post On X (East was known as Twitter), Microsoft AI CEO Mustafa Suleman posted about the My-DXO system. Calling it “a major step towards medical superintending”, he said that the AI system can solve some of the world’s most difficult medical cases with high accuracy and low cost compared to traditional clinical measures.
MAI-DXO simulates a virtual panel of physicians with various clinical approaches that collaborate to solve medical matters, the company said blog postThe orchestrator involves a multi-agentic system where a hypothesis provides, performs one test, provides two other checklists and stewardships, and challenges the final hypothesis.
My-DXO Workflow
Photo Credit: Microsoft
Once a hypothesis passes this panel, the AI system can either ask a question, request test, or diagnose if it seems that it seems to have enough information. If it recommends a test, it analyzes a cost to ensure that the overall cost remains appropriate. Interestingly, the system model is unknowable, which means that it can perform with any third-party AI model.
Microsoft claims that the system increases the clinical performance of each AI model tested. However, OPENAI’s O3 performed the best by correctly resolving 85.5 percent of the New England Journal of Medicine (NEJM) benchmark cases. The company said that the same cases were given to doctors from the US and the UK 21 practices, and they all had a clinical experience of five to 20 years. The accuracy of human doctors was 20 percent.
The MAI-DXO can be configured to operate within defined cost barriers, the company said. Once an input budget is added, the system examines the cost-to-board trades while taking clinical decisions. This helps only order the required tests in the AI system, rather than every possible test to control all causes of symptoms.
To assess the AI system, Microsoft also developed a new benchmark, called sequential diagnosis benchmark (SD bench). Unlike the specific medical benchmark tests asking multi-friendly questions, this test assesses the ability of the AI system to ask the correct questions and order the correct tests. It then evaluates answers by comparing them to the result published in Nejm.
In particular, the MAI-DXO has not yet been approved for clinical use, and it means initial research in developing AI capacity in clinical operations. Microsoft said that its AI system can only be approved for clinical use after rigorous safety testing, clinical verification and regulatory reviews.