Medicine can be a combination of art and science, but Microsoft has just demonstrated that much of both can be learned – by a bone.
The company reports in a study Published on the preprint site Arxiv that the AI-based medical program, the Microsoft AI Diagnostic Orchestrator (Mai-DXO), correctly diagnosed 85% of the cases in the New England Journal of Medicine. That is four times higher than the accuracy percentage of human doctors, who came up with the right diagnoses about 20% of the time.
The cases are part of the weekly series of the Journal, designed for doctors too stump: complicated, challenging scenarios where the diagnosis is not clear. Microsoft TOK about 300 of these Cases and Compared the Performance of its Mai-Dxo to that of 21 General-Practice Doctors in the US and UK in Order to Mimic the ITERATIVE WAY DOCTORS TYPOCLY SUCK CASSES -BE COLLECTING ITOSE, AND SOOKING ITOSE, ORDERING TESSE, ORDERING TESSO, Ordering. Results-microsoft’s team first created a stepwise decision-making benchmark process for each case study. As a result, both the doctors and the AI system were able to ask questions and make decisions about the following steps, such as ordering tests, based on the information they have learned at each step-comparable with a current diagram for decision-making, with subsequent questions and actions based on the previous one.
The 21 doctors were compared with a pooled set of off-the-shelf AI models that include Claude, Deepseek, Gemini, GPT, Grok and Llama. In order to further reflect the way in which human doctors approach such challenging cases, the Microsoft team also built an orchestrator: a virtual emulation of the sounding board of colleagues and consultation that doctors often search in complex cases.
In the real world, ordering medical tests costs money, so Microsoft followed the tests that the AI system and the human doctors ordered to see which method could do it cheaper.
Mai-Dxo not only performed much better than doctors in landing the correct diagnosis, but the AI-Bot could do this on average at 20% lower costs.
Read more: If Thimerosal is safe, why is it removed from vaccines?
“The four -time increase in accuracy had shown more than previous studies,” says Dr. Eric Topol, chairman of translational medicine and director and founder of the Scripps Research Translational Institute, who gave insights about the project. “Usually there is an absolute percentage difference of 10%, so this is a very big leap.” But what his attention really got were costs. “Not only was the AI more accurate, but it was much less expensive,” he says.
Mai-Dxo is still developing and is not yet available for use outside research. But the inclusion of such a model in medicine may lead to reductions of medical errors, which represents a significant share of health care costs and increasing the efficiency of human doctors – which in turn could lead to better results for patients.
“This is a surprising result,” says Mustafa Suleyman, CEO of Microsoft AI. “I think it gives us a clear line of sight to make the very best diagnostics of experts available to everyone in the world at an incredibly affordable price.”
A decade ago, when AI algorithms were first introduced in medicine, they were focused on binary tasks, says Suleyman, such as scanning images to detect tumors. “Nowadays, these models have flowing conversations of very high quality, they ask the right questions and sound in the right ways, which suggests the right tests and interventions at the right time,” he says.
Another advantage that an AI system can have is that it is free from many of the prejudices that are inherent in human experience. “We all have a confirmation advantage,” says Dr. Dominic King, vice -president of Microsoft AI. “Sometimes clinicians will see and think something:” I am sure this is just like the patient I saw last week. ” But ai thinks something else.
Read more: The surprising reason that national hospitals close
Mai-Dxo does not just spit out an answer. It shows its work, so that doctors may be able to study and investigate his reasoning process. “It is available for real -time supervision by the human doctor,” says Suleyman. “That is a level of transparency and visibility in the thinking process that we have not seen before.” That in turn could improve the educational and training that doctors receive to further increase the diagnostic accuracy and ultimately the results of the patient.
Yet some experts in the field of AI and medicine notice that Microsoft’s approach is not entirely new, because the diagnoses depend on the combined performance of multiple AI models. “In my mind they do not test an individual model that is optimized for health care,” says Keith Dreyer, Chief Data Science Officer at Massachusetts General Hospital and Brigham and Women’s Hospital Center for Clinical Data Science. “They test the concept of testing all the models that are there today and combine their decision -making together. That part is not surprising for me.”
Dreyer also points out that the results such systems are not necessarily brought closer to approved by regulatory authorities such as the US Food and Drug Administration, which still have not weighed whether such systems are medical devices or not.
Read more: What 105 blood tests get from a startup of health taught me
Microsoft is not the only company that strives for an AI-based medical program for diagnosing diseases. Google develops one Conversations based system To imitate the doctor-patient back and forth, to simulate the reasoning of human doctors in collecting information from patients and interpreting those symptoms to land in a diagnosis. In early tests, the system performed better than doctors in the accurate diagnosis of simulated case studies of the patient. In a 2024 test Similar to that Microsoft using case studies, the earlier version of Google system accurately diagnosed 59% of the cases, compared to the percentage of humans of 33%.
However, the real test will be how these AI systems perform in actual health systems. That is the next step to understand how AI could supplement or supplement the role of the doctor in diagnosing diseases. “It is impressive what they did,” says Topol. “But it does not change medical practice until they get it on the real medical highway.”
Topol hopes that the AI systems will be tested in various health systems, where doctors and the AI platform can be compared in a number of different and more typical cases. This would require a complete clinical study, as well as approval of regulatory authorities to ensure that no patients are exposed to damage by relying more on AI-based decision-making when providing their care. “We are very much on that trip to create the evidence base that is needed to support both clinics and patients to make a difference in their health,” says King.
If confirmed, such results can form the stage for introducing high -quality medical expertise in parts of the world that may not have access to important academic institutions or advanced healthcare. “My primary focus in the next five to 10 years is to ensure that everyone in the world has access to the very best medical advice of all species,” says Suleyman. “We are very enthusiastic about this.”
#Microsofts #doctors #diagnosing #diseases