Introduction: The rapid integration of artificial intelligence (AI) systems into healthcare, while improving diagnostic and therapeutic efficiency, raises serious concerns regarding the long-term impact on physicians’ professional competencies. A central risk is deskilling – the gradual and irreversible loss of clinical skills resulting from excessive reliance on automated systems.

Objective: To analyze the potential mechanisms of AI-induced deskilling, assess the feasibility of its objective diagnosis, and propose possible solutions to this problem.

Materials and Methods: This brief review examines latest scientific publications addressing the influence of AI on physicians’ professional activities. The analysis focuses on three key aspects of clinical practice: physical examination, diagnosis, and patient communication. Additionally, a specific empirical case is examined in detail, in which objective quantitative metrics were employed to assess skill degradation.

Results: The key mechanisms of AI’s negative impact were identified, including the degradation of manual skills, the «illusion of accuracy», «automation-induced bias», and the weakening of diagnostic reasoning, empathy, and moral competence. The evidence presented indicates that prolonged AI use can lead to a statistically significant decline in clinical performance, as substantiated by objective data such as a reduced adenoma detection rate during colonoscopy. Proposed solutions encompass the transformation of medical education, optimization of physician workflow, implementation of automated attention-monitoring systems, and a transition toward personalized competency maintenance programs.

Conclusion: Minimizing the risks of deskilling requires a systematic, rather than ad hoc, integration of AI into medicine, grounded in principles of ethics and pedagogical soundness. The central task is to create an environment in which technology serves to augment, rather than replace, human expertise, thereby preserving the balance between technological progress and the humanistic essence of medical practice.

Background. Currently, there are no algorithms capable of accounting for interactions of three or more concomitantly administered medications, which is necessary to reduce the risks of side effects associated with polypharmacy. To develop such an algorithm, we propose using a Bayesian network – a graphical probabilistic model based on a semantic graph constructed using information extracted from medical texts. However, constructing such a network requires analyzing a vast amount of available medical texts, which is extremely time-consuming and intellectually demanding for medical professionals, necessitating the automation of this process.

Objective. To compare various natural language processing architectures and deep learning technologies for the creation a tool for automatic analysis of medical text information on pharmacokinetics and pharmacodynamics, side effects, and interactions to create semantic graphs, subsequently used to build a Bayesian network for drug interaction analysis.

Methods. Natural language processing can be used to automatically tag medical instruction texts. In this study, deep learning models based on the BERT and T5 architectures were tested using the BIO format to extract named entities.

Results. BERT-based models showed the highest accuracy: RuBioBERT achieved an accuracy of 0.9569±0.0052, slightly outperforming ruBERT, which showed comparable results. It was found that pre-training on general texts provides an advantage in the initial stages, while specialized models such as RuBioBERT become more effective with an increasing number of training epochs. This is due to the fact that general texts provide a more universal language representation, while scientific medical articles have their own specific features which do not fully correspond to the structure of drug labels.

Conclusion. BERT models achieve near-expert-quality annotation, making them an effective tool for constructing semantic graphs of drug interactions. Automated NER annotation reduces expert workload and improves text processing accuracy. Therefore, the use of BERT models, such as ruBERT and RuBioBERT, is a promising method for automating medical text analysis and building platforms for assessing complex drug interactions.

Aim: to investigate the capabilities of artificial intelligence systems (AIS) in an inpatient setting, specifically in identifying signs of infiltration in chest radiographs (RG OGK).

Materials and methods. The initial data were chest X-rays (CXRs) in a frontal projection of 1,537 inpatients aged 18 to 102 years, obtained from the Unified Radiological Information System of the Unified Medical Information and Analytical Service (ERIS EMIAS). The presence of infiltration on the chest X-ray was determined by a radiologist (Radiologist 1) and an expert radiologist (Radiologist 2) with more than 5 years of experience. Radiologist 2 annotated the studies in two stages: at the first stage, only the chest X-rays were provided; at the second stage, in addition to the radiological data, access to information about the patient’s condition was provided.

Results. The accuracy rates for patients with a shifting state were 0.755 for AIS, 0.756 for Radiologist 1, and 0.856 for Radiologist 2, respectively. No statistically significant differences in accuracy were found between AIS and Radiologist 1 (p>0.05). When considering the scenario of using the AIS as a second opinion the accuracy for patients with a shifting state was 0.833, which is statistically significantly higher than the accuracy of Radiologist 1 and the AIS (p<0.05).

Conclusion. AIS can be used as a second opinion when detecting signs of infiltration based on RG OGK data in a hospital setting. However, since the diagnostic accuracy of the AIS is still insufficient for practical use, further research is needed to train algorithms using both visual data and electronic medical record data.

The aim of the study was to validate the Russian-language version of the SNAIL (Scale for the Assessment of Nonexperts’ AI literacy) questionnaire and evaluate its suitability for measuring AI literacy of healthcare professionals.

Materials and methods: online survey; two samples – healthcare professionals (n = 657 after filtering) and students in information-technology/mathematics programs (n=180). Psychometric analysis included Cronbach’s alpha for the subscales, within-group comparisons (paired-samples t-tests with Bonferroni correction), and between-group comparisons (Mann–Whitney U test).

Results: internal consistency was good to very good. Students significantly outperformed healthcare professionals on all subscales and on the total score. Within groups, healthcare professionals had more critical evaluation, more practical application, and more technical understanding; among students, technical understanding was lower than the other two subscales, whereas critical appraisal and practical application did not differ.

Conclusion: The Russian version of SNAIL questionnaire demonstrates reliability and discriminant validity and can be used to monitor healthcare workers’ AI literacy and plan educational interventions; a limitation is the self-report nature of the scale.

Breast cancer (BC) remains the leading cancer morbidity among women in Russia. Despite widespread screening, early-stage (0-II) incidence rates are low due to insufficient radiology staff, high consultation loads, and the combination of unsatisfactory mammograms.

Objective. To evaluate the effectiveness of artificial intelligence (AI) technology in interpreting mammography results during preventive medical examinations (PMEs) and early adult screenings (EAS) in reducing the detection rate of malignant breast tumors.

Materials and methods. A multicenter observational study with historical control was conducted at healthcare facilities in six constituent entities of the Russian Federation (Bryansk, Saratov, Novosibirsk, and Sakhalin Oblasts, Kamchatka Krai, and Khabarovsk Krai). During the study period, 677,652 mammography examinations were performed in the pilot regions under the Primary Medical Association (PMA) and the Department of State Health Supervision (DOGVN), of which 247,514 (36.6%) were performed using the Third Opinion AI analysis service (RZN 2022/16534). The share of examinations submitted for AI processing ranged from 18.0% (Khabarovsk Krai) to 79.6% (Kamchatka Krai). The comparison group included 553,011 examinations from the same regions in 2023. Correlation analysis was performed using the Spearman rank correlation method.

Results. The primary BC incidence rate in 6 regions increased from 165.5 (2023) to 270.3 (2024-2025) per 100,000 mammograms (+63.4%). The incidence of stages 0-II increased from 144.5 to 212.4 (+47.0%). The share of BI-RADS 4-5 was 23.6% with the use of AI versus 10.8% with routine screening. A statistically significant rank correlation was found between AI coverage and the share of BI-RADS 4-5 (ρ = 0.886; p = 0.019; n = 6).

Conclusion. The introduction of AI in the description of mammography results during primary medical examinations/preclinical screening is associated with an increase in the primary BC detection rate and the number of cases diagnosed at early stages. Medical effectiveness is consistent with data from large randomized trials. Scaling up requires organizational support.

The study presents the development of a hardware-software system for neurorehabilitation based on motor imagery, implementing the full cycle of brain-computer interface (BCI) operation – from EEG acquisition and preprocessing to classification, multimodal feedback control, and generation of clinically oriented reports.

Materials and Methods: The system architecture includes patient and scenario databases, a data stream synchronization module, automatic removal of oculomotor artifacts using Independent Component Analysis (ICA), a classifier based on Riemannian geometry of covariance matrices, and an adaptive feature alignment mechanism to reduce inter-individual variability.

Results: The developed system ensures reliable real-time performance and generates reports including neurophysiological parameters (spectral power in delta, theta, alpha, and beta bands, ERD, lateralization index) and behavioral characteristics (motor imagery formation speed, number of successful imagery attempts). The system provides objective monitoring of patient’s condition and recovery dynamics, including assessment of changes in background rhythmic activity and functional responses of the sensorimotor cortex.

Conclusions: The proposed solution combines state-of-the-art algorithmic methods with clinical applicability, making it a promising tool for neurorehabilitation and cognitive-motor training.

IThe article examines the development algorithm and architecture of a prototype information-analytical decision support system (IADSS) integrating data on a network of medical organizations, their resources, and patient flows. The system is designed to operate both in routine mode and in emergency situations, providing forecasting, optimization of resource allocation, and coordination between departments and medical organizations.

The IADSS prototype was developed using a modular approach and a service-oriented architecture. Geoinformation services, modern programming languages, frameworks, analytical libraries, and visualization tools were used in the development.

The architecture includes a geoinformation module for displaying and analyzing medical infrastructure, a BPMN business process editor, a simulation and optimization modeling module, calculators for computing integral sustainability indices, as well as tools for reporting, monitoring key performance indicators (KPI), managing resource inventories and information security.

The implemented IADSS prototype combines more than ten interconnected microservices, ensuring flexibility, scalability, and the ability to integrate with external information and analytical systems. The viability of the architectural approach was verfied through functional testing on synthetic datasets simulating biological, technogenic (man-made), and geopolitical threats scenarios. The comprehensive integration of geoinformation technologies, business process modeling, analytics, and optimization tools improves the preparedness of healthcare organizations to respond to emergencies, minimizes the impact of external threats, and reduces decision-making time. The proposed architecture can serve as the basis for the further development of regional decision support systems in healthcare, enhancing the system’s resilience and operational efficiency in the face of global challenges.