The first AI service that analyzes X-rays starts operating in Moscow healthcare system

The first AI service that analyzes X-rays starts operating in Moscow healthcare system

Artificial intelligence in the capital's medical organizations has been working for the second month since May’2020 and now, in addition to computed tomograms of the chest organs, it will also begin to analyze X-rays. To do this, the Care Mentor AI software was integrated into the Unified Radiological Information Service (URIS), to which the diagnostic equipment of Moscow is connected, which analyzes chest radiographs for the presence of various pathologies, including such socially significant ones as lung cancer, tuberculosis, pneumonia.

Care Mentor AI is a Russian developer of computer vision services for the medical images analysis in based on neural network algorithms. Its product - the service "X-ray screening of the chest organs" - has recently been successfully integrated into the URIS working circuit and connected to 17 medical organizations in Moscow with new healthcare organizations added each day.

This service is screening for lung pathology. It analyzes X-ray images and within a few seconds determines the presence or absence of pathological changes in the chest organs in eight groups of signs (including socially significant diseases). He also conducts triage, ranking studies by how serious the pathology is and how quickly it needs to be addressed.

AI models are trained on a large (10k +) volume of X-ray images, marked by professional radiologists. Both classification models and segmentation models were used. A separate neural network was trained for each group of radiological signs of pathology. The predictions of these neural networks are combined using a second-level model.

“The start of work with medical organizations in Moscow is an important step for our team, which allowed us to participate in practical assistance to a large number of radiologists and patients. Of course, the serious and multifaceted challenge associated with the spread of the coronavirus made us, as an AI development company for medicine, look at our services in a different way.
For example, we trained our neural network system to detect signs of pneumonia on chest X-ray images, which will undoubtedly expand its functionality and allow it to be used in mass examinations of patients during periods of seasonal outbreaks of respiratory diseases,” says Pavel Roytberg, co-founder of Care Mentor AI.

The integration of services based on computer vision algorithms into the city diagnostics system is taking place as part of a large-scale Moscow experiment launched by the Moscow City Health Department and the Center for Diagnostics and Telemedicine in February 2020. Its goal is to comprehensively investigate the capabilities to use of innovative computer vision technologies for medical image analysis and subsequent applicability in the healthcare system of Moscow.

The Center for Diagnostics and Telemedicine notes that before accessing ERIS, services must undergo functional testing and calibration. During functional testing the compliance of the AI ​​service with the parameters declared by the developers is assessed.

“The stability of each function is checked many times during testing. Further, according to 11 main criteria, it is analyzed to what extent the AI ​​service meets the declared parameters. Based on the results of functional testing, critical or non-critical remarks are formed. If there are no critical remarks or they can be promptly eliminated, then the service can be implemented into the main, working contour of the ERIS.

The Care Mentor AI service has successfully passed the listed stages of testing and became the first in X-ray studies, ”says Sergey Morozov, Chief Specialist for Radiological and Instrumental Diagnostics of Moscow, Director of the Diagnostic and Telemedicine Center of the Moscow Health Department.

The Care Mentor AI product was one of the first to enter the "working" stage of the experiment, and another service of the company for detection the volume and severity of lung damage with COVID-19 is at the first stage of the Experiment and is approved to move on to technical integration.