Google's AMIE and German MIRA AI Match or Beat Doctors in Diagnosis
The results mark a step toward AI tools that could assist in clinical decision-making, though researchers caution the test conditions do not yet reflect real-world clinical settings.
Key Facts
- On June 17, 2026, two research papers in Nature reported that AI tools Google's AMIE and German-developed MIRA matched or exceeded human doctors in diagnosis and treatment.
- MIRA achieved 87.1% diagnostic accuracy across eight diseases including appendicitis and pulmonary embolism, outperforming a six-doctor panel's 78.1% accuracy.
- MIRA was tested on over 500 emergency department cases, retrieving patient data from electronic medical records and selecting from 85,000 options.
- AMIE matched 21 doctors in reasoning ability and exceeded them in guideline adherence and medication decisions for difficult cases.
Reporting from 1 source: GIGAZINE.
Two AI medical tools, Google's AMIE and the German-developed MIRA, demonstrated diagnostic and treatment performance equal to or better than human doctors in studies published in Nature. MIRA achieved 87.1% diagnostic accuracy across eight diseases, outperforming a panel of six doctors. AMIE matched doctors in reasoning and exceeded them in guideline adherence and medication decisions for difficult cases.
On June 17, 2026, two research papers published in Nature reported that AI medical tools Google's AMIE and the German-developed MIRA matched or exceeded human doctors in diagnostic accuracy and treatment decisions. MIRA, which retrieves patient data from electronic medical records and selects from 85,000 options including tests, medications, and procedures, was tested on over 500 emergency department cases. It achieved 87.1% accuracy for eight diseases such as appendicitis and pulmonary embolism, beating the 78.1% accuracy of a six-doctor panel. AMIE, which makes treatment plans based on patient-provided data, matched 21 doctors in reasoning ability and adhered more closely to clinical guidelines, also outperforming experts on medication decisions for difficult cases.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.