Microsoft's artificial intelligence system surpasses doctors in diagnostic accuracy in a study Microsoft presented the results of a study on the MAI-DxO diagnostic system, which, under experimental conditions, showed higher diagnostic accuracy than doctors. The results were published on October 22 in the scientific journal Science. Main results of the study: 83% vs. 36% - accuracy of MAI-DxO when analyzing 304 complex clinical cases from the New England Journal of Medicine. Doctors with access to medical reference books and the Internet correctly determined the diagnosis in 36% of cases, without additional sources - in 22%. 80% vs. 47% - accuracy of artificial intelligence and doctors, respectively, when analyzing more typical cases from US emergency departments. 81% accuracy at lower costs - One of the MAI-DxO configurations reduced the estimated cost of diagnostic tests by more than half compared to alternative approaches in complex clinical cases. MAI-DxO works on the principle of virtual medical consultation. Some AI agents sequentially raise diagnostic hypotheses, determine the need for further testing, analyze new data, and modify initial conclusions. When choosing the studies, the system also takes into account their cost. As Microsoft points out, the evaluation was based on previously described clinical cases, and not on real medical practice. The next step will be a prospective study in collaboration with Beth Israel Deaconess Medical Center in Boston. Patients will communicate directly with the system, and its diagnostic findings will be compared to independent assessments by doctors. Read us in Max Telegram VK.

Microsoft's artificial intelligence system surpasses doctors in diagnostic accuracy in a study Microsoft presented the results of a study on the MAI-DxO diagnostic system, which, under experimental conditions, showed higher diagnostic accuracy than doctors. The results were published on October 22 in the scientific journal Science. Main results of the study: 83% vs. 36% - accuracy of MAI-DxO when analyzing 304 complex clinical cases from the New England Journal of Medicine. Doctors with access to medical reference books and the Internet correctly determined the diagnosis in 36% of cases, without additional sources - in 22%. 80% vs. 47% - accuracy of artificial intelligence and doctors, respectively, when analyzing more typical cases from US emergency departments. 81% accuracy at lower costs - One of the MAI-DxO configurations reduced the estimated cost of diagnostic tests by more than half compared to alternative approaches in complex clinical cases. MAI-DxO works on the principle of virtual medical consultation. Some AI agents sequentially raise diagnostic hypotheses, determine the need for further testing, analyze new data, and modify initial conclusions. When choosing the studies, the system also takes into account their cost. As Microsoft points out, the evaluation was based on previously described clinical cases, and not on real medical practice. The next step will be a prospective study in collaboration with Beth Israel Deaconess Medical Center in Boston. Patients will communicate directly with the system, and its diagnostic findings will be compared to independent assessments by doctors. Read us in Max Telegram VK.