Currently, information technologies are being actively introduced in the healthcare of the Russian Federation. The share of state and municipal medical organizations that have implemented various medical information systems increased from 3.9 % in 2007 to 91 % in 2021. One of the key tasks of informatization is the introduction of electronic medical records (EMRs), which accumulate large amounts of Real-World Data (RWD). Despite the importance of EHR as a source of RWD, they have a number of shortcomings, such as the decentralized nature of database management systems, unstructured information storage, etc. The article describes the sequential processes for collecting high-quality RWD based on EHR, including the use of artificial intelligence technologies, for the purposes of scientific research, the creation of decision support systems, statistical analysis, etc. The basis of the proposed methodology is the centralized collection of information from EMR in the so-called data lakes, where as much as possible of raw data on the patient is accumulated and subsequent extraction of data from unstructured records through natural language processing (NLP) models. The proposed technology, subject to continuous improvement, will provide a correct and comprehensive solution for the skilful understanding of any text from any medical record.

During the pandemic of novel coronavirus infection (COVID-19), computed tomography (CT) showed its effectiveness in diagnosis of coronavirus infection. However, ionizing radiation during CT studies causes concern for patients who require dynamic observation, as well as for examination of children and young people. For this retrospective study, we included 15 suspected for COVID-19 patients who were hospitalized in April 2020, Russia. There were 4 adults with positive polymerase chain reaction (PCR) test for COVID-19. All patients underwent magnetic resonance imaging (MRI) examinations using MR-LUND PROTOCOL: Single-shot Fast Spin Echo (SSFSE), LAVA 3D and IDEAL 3D, Echo-planar imaging (EPI) diffusion-weighted imaging (DWI) and Fast Spin Echo (FSE) T2 weighted imaging (T2WI). On T2WI changes were identified in 9 (60,0%) patients, on DWI – in 5 (33,3%) patients. In 5 (33,3%) patients lesions of the parenchyma were visualized on T2WI and DWI simultaneously. At the same time, 4 (26.7%) patients had changes in lung tissue only on T2WI. (P(McNemar) = 0,125; OR = 0,00 (95%); kappa = 0,500). In those patients who had CT scan, the changes were comparable to MRI. The results showed that in case of CT is not available, it is advisable to conduct a chest MRI for patients with suspected or confirmed COVID-19. Considering that T2WI is a fluid-sensitive sequence, if imaging for the lung infiltration is required, we can recommend the abbreviated MRI protocol consisting of T2 and T1 WI. These data may be applicable for interpreting other studies, such as thoracic spine MRI, detecting signs of viral pneumonia of asymptomatic patients. MRI can detect features of viral pneumonia.

The article analyzes ethical issues inherent to different life-cycle stages of artificial intelligence systems and provides up-to-date information about global and domestic trends in this area. The international and national experience concerning ethical issues of artificial intelligence systems use in healthcare is described. In addition, the international and national strategies for the development of artificial intelligence in healthcare are analyzed focusing on the national development; and the main trends, similarities, and differences between strategies are identified. Furthermore, ethical components of the process of clinical trials to evaluate the safety and efficacy of artificial intelligence systems in Russia are described. Domestic state-of-the-art and globally unique experience in technical regulation of artificial intelligence systems are shown on unification papers and standardization of requirements for the development, testing, and operation of artificial intelligence systems in healthcare are presented, and unparalleled Russian experience in terms of certification requirements for artificial intelligence-based medical devices is demonstrated. The article also summarizes the main conclusions and emphasizes the importance of a strong successful healthcare system based on artificial intelligence technologies that build trust and compliance with ethical standards.

BACKGROUND: Pathological low-energy (LE) vertebral compression fractures (VFs) are common complications of osteoporosis and predictors of subsequent LE fractures. In 84% of cases, VFs are not reported on chest CT (CCT), which calls for the development of an artificial intelligence-based (AI) assistant that would help radiology specialists to improve the diagnosis of osteoporosis complications and prevent new LE fractures.AIMS: To develop an AI model for automated diagnosis of compression fractures of the thoracic spine based on chest CT images.MATERIALS AND METHODS: Between September 2019 and May 2020 the authors performed a retrospective sampling study of ССТ images. The 160 of results were selected and anonymized. The data was labeled by seven readers. Using the morphometric analysis, the investigators received the following metric data: ventral, medial and dorsal dimensions. This was followed by a semiquantitative assessment of VFs degree. The data was used to develop the Comprise-G AI mode based on CNN, which subsequently measured the size of the vertebral bodies and then calculates the compression degree. The model was evaluated with the ROC curve analysis and by calculating sensitivity and specificity values.RESULTS: Formed data consist of 160 patients (a training group - 100 patients; a test group - 60 patients). The total of 2,066 vertebrae was annotated. When detecting Grade 2 and 3 maximum VFs in patients the Comprise-G model demonstrated sensitivity - 90,7%, specificity - 90,7%, AUC ROC - 0.974 on the 5-FOLD cross-validation data of the training dataset; on the test data - sensitivity - 83,2%, specificity - 90,0%, AUC ROC - 0.956; in vertebrae demonstrated sensitivity - 91,5%, specificity - 95,2%, AUC ROC - 0.981 on the cross-validation data; for the test data sensitivity - 79,3%, specificity - 98,7%, AUC ROC - 0.978.CONCLUSIONS: The Comprise-G model demonstrated high diagnostic capabilities in detecting the VFs on CCT images and can be recommended for further validation.