{"id":381895,"date":"2026-10-23T03:16:00","date_gmt":"2026-10-23T10:16:00","guid":{"rendered":"https:\/\/cms-articles.softonic.io\/en\/?p=381895"},"modified":"2026-10-23T03:16:07","modified_gmt":"2026-10-23T10:16:07","slug":"microsoft-publishes-science-study-its-ai-diagnostic-ai-outperforms-doctors","status":"publish","type":"post","link":"https:\/\/cms-articles.softonic.io\/en\/microsoft-publishes-science-study-its-ai-diagnostic-ai-outperforms-doctors\/","title":{"rendered":"Microsoft publishes Science study: its AI diagnostic AI outperforms doctors"},"content":{"rendered":"

Microsoft published a Science paper today on MAI-DxO, its AI diagnostic system, with new results on how the software handled simulated medical diagnosis.<\/p>\n

\r\n
\r\n
\r\n
\r\n \"Microsoft\r\n <\/div>\r\n
\r\n Microsoft Copilot<\/span>\r\n Download<\/a>\r\n <\/div>\r\n
\r\n <\/path><\/path><\/text><\/svg>\r\n <\/div>\r\n <\/div>\r\n
\r\n <\/span>\r\n <\/div>\r\n
\r\n \r\n <\/div>\r\n <\/a>\r\n <\/div>\r\n<\/div>\n\n\n\n

Across 304<\/strong> difficult New England Journal of Medicine cases, MAI-DxO reached 83.2% accuracy, according to Microsoft\u2019s paper, compared with 36.1% for generalist doctors using textbooks and the internet; Microsoft adds that, in some more resource-limited physician comparisons, the number was closer to 22%. In another set of 325 emergency admissions, MAI-DxO hit 80.3% versus 46.8% in the same study, about 1.7 times better. The way it got there matters. Rather than making a one-shot guess, the system worked through each case step by step, deciding what to ask next, which tests to order, and when to revise the diagnosis.<\/p>\n\n

Microsoft says MAI-DxO also chose fewer tests<\/strong> at a lower simulated cost. The strongest results came from GPT-5 and GPT-5.5, though Microsoft says the system can run on models from Google, Anthropic, Meta, and others. There are some clear limits to keep in mind: the study was text-only, the physician comparison used generalists rather than specialists, and neither side had access to physical exams or imaging.<\/p>\n\n

If you follow AI in healthcare<\/a>, the paper is worth reading. It still doesn\u2019t prove that software can replace a doctor in the exam room. Workflow, liability, trust, and value are all still unresolved.<\/p>\n\n

The Science paper is available now. Microsoft says MAI-DxO isn\u2019t treating patients, and the next round of testing is planned with Beth Israel Deaconess for emergency triage and waiting-room support, as Microsoft open-sources SDBench<\/strong>.<\/p>\n

\r\n
\r\n
\r\n
\r\n \"ChatGPT\"\r\n <\/div>\r\n
\r\n ChatGPT<\/span>\r\n Download<\/a>\r\n <\/div>\r\n
\r\n <\/path><\/path><\/text><\/svg>\r\n <\/div>\r\n <\/div>\r\n
\r\n <\/span>\r\n <\/div>\r\n
\r\n \r\n <\/div>\r\n <\/a>\r\n <\/div>\r\n<\/div>\n\n","protected":false},"excerpt":{"rendered":"

Microsoft published a Science paper today on MAI-DxO, its AI diagnostic system, with new results on how the software handled simulated medical diagnosis. Across 304 difficult New England Journal of Medicine cases, MAI-DxO reached 83.2% accuracy, according to Microsoft\u2019s paper, compared with 36.1% for generalist doctors using textbooks and the internet; Microsoft adds that, in … Continue reading “Microsoft publishes Science study: its AI diagnostic AI outperforms doctors”<\/span><\/a><\/p>\n","protected":false},"author":9349,"featured_media":381894,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","wpcf-pageviews":0},"categories":[1015],"tags":[],"usertag":[],"vertical":[],"content-category":[6771],"class_list":["post-381895","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","content-category-ai"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t