The Russian neural network has surpassed the accuracy of Microsoft's AI diagnostics

The medical AI assistant, developed by the AIRI Institute's specialists based on the GigaChat neural network model, achieved 93% accuracy in detecting diseases during the test. Earlier, a similar system from Microsoft achieved an 85% rate in a similar experiment, the creators of the domestic product told Izvestia. The program needed to diagnose pathologies using real-world case studies from the respected New England Journal of Medicine.

In concert with sounds: AI will customize children's hearing implants in minutes

How a neural network will help speed up the adaptation of young patients with hearing loss

The Russian development was acting blindly. First, she received primary information about the patient — gender, age, and symptoms. At the next stage, the AI requested the results of tests, visualization, and consultations step by step in order to finally make its diagnosis. On average, three actions were enough for the model to do this.: collecting data from a patient — a comment or a question from a doctor — the patient's response.

All the analyzed cases were divided into three types of complexity. A Russian-developed neural network has successfully identified rare diseases such as Whipple's disease, aceruloplasminemia and methemoglobinemia.

"Today, multi-agent systems are able to confidently search for rare, masked pathologies beyond the typical algorithms of the admission department. The AI assistant showed that it was able to rebuild a hypothesis in time, rejecting the most likely but false path, and did it faster than an expert observer with 15 years of experience in medicine expected," said Ivan Oseledets, Director General of the AIRI Institute.

The Russian AI assistant is currently undergoing testing. His help has already been used more than 160 thousand times.

Read more in the exclusive Izvestia article:

Dr. AI-hurts: the Russian neural network has surpassed the accuracy of Microsoft's AI diagnostics