Microsoft’s AI Is Better Than Doctors at Diagnosing Disease

Medicine may be a mixture of art and science, but Microsoft has just shown that a lot can be learned from the two – through robot.

Company reports in a Ticket Posted on Preprint Arxiv The Microsoft Ai Diagnostic Orthrator (MAI-DXO) program was properly diagnosed 85 % of the cases described in New England Magazine for Medicine. This is four times higher than the accuracy of human doctors, who have reached correct diagnoses about 20 % of the time.

Cases are part of the weekly magazine series designed for doctors ’doctors: complex and challenge scenarios where the diagnosis is not clear. Microsoft took about 300 of these cases and compared the performance of Mai-DXO with 21 general practitioners in the United States and the United Kingdom to simulate the repetitive method of doctors usually such cases-by collecting information, analyzing them, arranging tests, and then making decisions based on those results-the Microsoft team first created by creating a step-by-making decisions to review everything. This allowed both doctors and artificial intelligence system to ask questions and make decisions about the following steps, such as requesting tests, based on the information they learned in each step-similar to the flow plan to make decisions, with questions and subsequent measures based on the information collected from that previous.

The 21st doctors were compared to a group of ready -made artificial intelligence models that included Claude, Depsic, Gemini, GPT, Grok, and Llama. For more unlike the way human doctors deal with such difficult cases, the Microsoft team has also built an orchestra: a virtual intercourse with the Perception Council for colleagues and consultations that doctors are looking for in complex cases.

In the real world, the request for medical tests costs funds, so Microsoft has tracked the tests that ordered the artificial intelligence system and human doctors knowing the way it can get cheaply.

Not only did the Mai-DXO outperformed doctors in the proper diagnosis, but artificial intelligence robot was able to do this at a 20 % lower cost on average.

Read more: If Thimosal is safe, then why is it removed from the vaccines?

“The increase that precedes the four accuracy in the accuracy is more than the previous studies,” says Dr. Eric Topol, head of medicine, director and founder of the Scripps Institute for Translation Research, who presented an insight into the project. “Most of the time, there is a 10 % absolute percentage, so this is a really big jump.” But what really caught his attention is the cost. “Amnesty International was not only more accurate, but it was much less expensive,” he says.

The mai-dxo is still under development and is not available for use outside the search yet. But merging such a model into medicine can lead to discounts in medical errors, which represent a large share of health care costs, and increasing the efficiency of human doctors – which in turn can lead to better results for patients.

“This is an amazing result,” said Microsoft Ai. “I think it gives us a clear line to provide the best diagnosis of experts for all in the world at an incredible price point.”

A decade ago, when Amnesty International’s algorithms were presented for the first time in medicine, they focused on bilateral tasks, says Suleiman, such as wiping pictures to detect tumors. He says: “Today, these models are taking place fluently in very high quality, asking the correct questions and investigating the correct methods, indicating the correct test and interventions in a timely manner,” he says.

Another feature that the artificial intelligence system may enjoy is that it is free of many biases in human experience. “We all have a confirmation bias,” says Dr. Dominic King, Vice President of Microsoft Ai. “Sometimes doctors will see something and think,” I am sure that this is like the patient I have seen last week. “But artificial intelligence thinks a little differently.”

Read more: The sudden cause of rural hospitals is closed

Mai-dxo does not spit an answer. It shows his work, so that doctors can study and check the thinking process. Suleiman says: “It is available to supervise in an actual time by the human doctor.” “This is a level of transparency and vision in the thinking process that we have not seen before.” This, in turn, can improve education and training doctors to increase the accuracy of the diagnosis and the patient’s results in the end.

However, some experts in the field of artificial intelligence and medicine note that the Microsoft approach is not completely new, since its diagnoses depend on the common performance of multiple AI models. “In my opinion, they do not test any individual model that has been improved for health care,” says Keith Drayer, chief data science employee at Massachusetts General Hospital and Women’s Hospital Center for Clinical Data Science. “They are testing the concept of testing all models there today and combining decisions together. This part is not surprising.”

Dreyer also indicates that the results do not necessarily bring such systems to be approved by organizational agencies such as the American Food and Drug Administration, which still weigh whether these systems are medical devices or not.

Read more: What taught me to get 105 blood tests from a healthy starter company

Microsoft is not the only company to follow a medical -based medical program to diagnose the disease. Google develops a Conversation To simulate the doctor and the patient with the background, the logic of human doctors mimics information from patients and interprets these symptoms on the decline in the diagnosis. In early tests, the system outperformed doctors in diagnosing the patient’s condition studies accurately. In 2024 exam Similar to one of Microsoft conducted with case studies, the previous version of the Google system was diagnosed with a resolution of 59 % of cases, compared to the rate of 33 % human doctors.