Bair Tuchinov, Viktor Suvorov, and Lyu Min Shao Khue, a team of researchers from the NSU Mechanics and Mathematics Department’s (MMD) Laboratory of Streaming Data Analytics and Machine Learning, are expanding artificial intelligence applications in the field of medical diagnostics. They have already developed a software module for recognizing and identifying brain tumors on MRI images. Research work is underway for a project related to identifying foci of multiple sclerosis together with the Multiple Sclerosis Center and the International Tomography Center SB RAS.
A new Laboratory initiative is aimed at developing scalable domestic artificial intelligence technologies capable of recognizing metastases and tuberculosis. This project will develop computer vision algorithms for diagnosing neuro-oncological and infectious diseases (detection, segmentation, classification, and tracking in space and time) using data from a set of computed tomography and radiography techniques. This work is being supported by the Priority 2030 Program.
Other project partners are the Novosibirsk Tuberculosis Research Institute, the Federal Center of Neurosurgery, Research Institute of Clinical and Experimental Lymphology of the Siberian Branch of the Russian Academy of Medical Sciences, V. Zelman Institute of Medicine and Psychology, as well as manufacturers of X-ray machines and MRI systems.
Evgeniy Pavlovsky, Laboratory Head, described the importance and trajectory of the project,
The problem of tuberculosis is acute in Siberia and the Far East. Other serious lung pathologies are no less relevant. Our project started in July of this year and is still at the initial stage of implementation. We already had preliminary developments and now we are working with the Novosibirsk Tuberculosis Research Institute, helping the specialists prepare data for training algorithms. Together we are creating the database necessary for this research. We established cooperation with manufacturers of X-ray machines and diagnostic magnetic resonance imaging systems. Several companies are interested in using the results in their equipment. By interacting with them, we are defining the contours of the future product for radiologists and neurologists as well as for the technicians who will configure the equipment. These contours are already visible. At first, our Laboratory considered the option of a software module integrated into the equipment, now we are thinking about creating software that could be used in the existing equipment at any medical and diagnostic medical institution, This will contribute to its widespread use.
Another important task is to teach artificial intelligence to recognize and segment areas of the lungs, ribs, and vertebrae in MRI images and X-rays, as well as to explore clinically significant signs and correlate them with various diseases. Laboratory specialists want to integrate all these tasks into a comprehensive solution so that artificial intelligence performs routine tasks including distinguishing healthy lungs from those affected by disease, finding the localization of pathological formations, outlining the contours of foci of disease and determining their location relative to the ribs and vertebrae, “highlighting” areas of special attention, and “suggesting” a possible diagnosis based on the experience of the many doctors on which it was trained.
Pavlovsky continued,
Our software module will not replace a doctor. We believe that only a doctor should make diagnoses and prescribe treatment, but artificial intelligence will become a useful “assistant” that will provide information, make the necessary focus, and highlight significant points. The doctor’s role in the process is to rely on his knowledge, skills, experience, and intuition.
The researchers plan to complete the project and receive a ready-made software module at the end of December this year. They are currently exploring available foreign data sets and adapting them to domestic ones. The first results have been obtained and have been assessed as positive by researchers and of interest for further work. During the testing process, the researchers presented their findings on this project, as well as on projects implemented earlier (specifically, on identifying and classifying types of brain tumors using artificial intelligence) at the Golden Valley Research and Production Forum that took place at NSU November 1-2.
