Moscow Department of Health and Moscow State University: Method of quality control of medical tomographs

The main articles are:

• Neural networks (neural networks)
• Artificial Intelligence in Medicine
• Machine Learning

2022: Presentation of the method of quality control of medical tomographs using the neural network

Specialists Diagnostic and Telemedicine Center , together with mathematicians MSU , have developed a quality control method medical tomographs that will allow timely detection of malfunctions of devices MRI in automatic mode. This was announced on December 9, 2022 by the Center for Diagnostics and Telemedicine DZM.

For December 2022, many digital services help Moscow radiologists, but we are always looking for opportunities to make radiation diagnostics even more efficient. To do this, the Center's specialists proposed an MRI control technique for clinical images and trained an algorithm developed by colleagues from Moscow State University to automate the process. This will allow you to quickly identify tomographs requiring additional attention from technical specialists and, as a result, reduce the duration of downtime and the cost of repair. The system still needs additional training and testing, but the existing results indicate the feasibility of implementation. This approach in the future can improve the quality of radiation diagnostics of the capital, - said Yuri Vasiliev, director of the Center for Diagnostics and Telemedicine of the DZM, chief specialist of Moscow in radiation and instrumental diagnostics.

Magnetic resonance imaging is a highly accurate method of three-dimensional imaging of internal organs without harmful ionizing radiation, increasingly used by radiologists to accurately make diagnoses. The presented method of quality control of medical tomographs was developed to avoid breakage and downtime of equipment. It is based on machine learning technology. To configure the neural network model, a sample of MRI images from various devices is collected, for which the quality control result is accurately known - whether the device is healthy or not. The algorithm is trained to distinguish between images from serviceable and faulty devices. Experimental evaluation on the data showed the superiority of the developed method over analogues in accuracy.

Such technology has a number of advantages. Firstly, the time of the X-ray absorber is saved, which needs to manually assess the quality of the devices. This procedure requires special training, and also takes some time. Secondly, the regularity of control is guaranteed. Quality control of equipment operation shall be carried out daily, in extreme cases weekly. Automatic image quality control can be performed in 24/7 format. Analyzing one three-dimensional image takes less than a second, so after the study, the system will immediately mark "suspicious" images. Further, the staff will be able to analyze the information received and, if necessary, call a technical team to repair or replace equipment, - said Olga Senyukova, associate professor of the Department of Intellectual Information Technologies, Faculty of Computational Mathematics and Cybernetics, Moscow State University, Ph.D.

The Diagnostic and Telemedicine Center provides advisory support on the safety of MRI rooms at the design stage and within the framework of acceptance and periodic tests, preparation of methodological and informational materials, as well as personnel training.