Aim. To evaluate the potential of artificial intelligence (AI) technologies as a "second pilot" tool for phenotyping elderly patients with atherosclerosis in various vascular systems and to study the clinical efficacy of combination lipid-lowering therapy (rosuvastatin + ezetimibe) in this population.

Material and methods. This prospective cohort study included 739 patients over 60 years of age (median, 67 [63; 72] years; women, 52,8%) with verified atherosclerosis (coronary, cerebral, or peripheral arteries). Using stratified randomization, patients were divided into two following groups: the main group (n=371) received combination therapy (rosuvastatin 20 mg + ezetimibe 10 mg/day) for 6 months; the control group (n=368) received standard lipid-lowering therapy (rosuvastatin 20 mg/day) in accordance with clinical guidelines. A machine learning algorithm (gradient boosting with SHAP-based interpretation) integrated into the clinical decision support system was used for patient phenotyping. Lipid profile, echocardiographic parameters, and changes in comorbid pathology were assessed at the baseline visit and after 3and 6-month follow-up.

Results. The AI algorithm allowed us to identify four following clinical phenotypes of patients with atherosclerosis: "coronary" (38,2%), "cerebrovascular" (27,6%), "multifocal" (19,1%), and "metabolic" (15,1%). In the study group, significant lipid profile improvement was recorded by the 6th month of follow-up as follows: a 54,2% decrease in low-density lipoprotein cholesterol (LDL-C) (p<0,001 vs control) and an 18,4% increase in HDL-C (p<0,01). Achievement of the target LDL-C level <1,4 mmol/L was observed in 67,4% of patients in the study group compared to 41,2% in the control group (p<0,001). Echocardiography data showed a 5,2% increase in left ventricular ejection fraction (p<0,01) and diastolic function improvement in the study group. The greatest treatment efficacy was observed in patients with the "coronary" and "multifocal" phenotypes.

Conclusion. The use of AI as a "second pilot" allows for therapeutic strategy personalization in elderly patients with atherosclerosis. Combination therapy with rosuvastatin and ezetimibe demonstrates high efficacy in achieving target lipid levels and positive dynamics in echocardiographic parameters, with the greatest effect in certain clinical phenotypes.

Aim. To evaluate the effectiveness of modern ensemble machine learning (ML) models in predicting in-hospital mortality in patients with acute coronary syndrome (ACS) compared to the traditional GRACE clinical score.

Material and methods. This retrospective study included anonymized data from 14420 patients with ACS admitted to a cardiology hospital. For each case, 28 predictors were analyzed. Following machine learning algorithms were trained using this data: logistic regression, decision trees, random forest, and gradient boosting (XGBoost, CatBoost). The area under the receiver operating characteristic (AUC-ROC), recall, and F1 score were used to assess the model quality. The results of the best model were compared with the GRACE risk score. SHAP analysis was used to interpret the algorithm logic.

Results. Inhospital mortality was 6,03% (n=804). Gradient boosting algorithms demonstrated the highest predictive capability. The CatBoost model emerged as the top performer, achieving an AUC-ROC of 0,961, significantly outperforming the GRACE score (AUC-ROC 0,919) on the test set. SHAP analysis revealed that the greatest contributors to the model’s prediction were history of dyslipidemia, left ventricular ejection fraction, Killip class of acute heart failure, age, and systolic blood pressure levels. The model successfully identified hidden nonlinear clinical patterns, including the paradoxical protective effect of previously diagnosed dyslipidemia and the critical prognostic significance of missing medical history data upon admission.

Conclusion. ML methods, particularly the CatBoost algorithm, provide higher accuracy in predicting in-hospital mortality in ACS compared to conventional scores. The ability of these algorithms to account for complex nonlinear relationships between clinical indicators makes them a promising foundation for developing accurate clinical decision support systems.

Objective. To develop a multivariate predictive model for the early detection of acute coronary occlusion (ACO) in patients with non-ST-elevation acute coronary syndrome (NSTE-ACS) using machine learning methods and a set of predictors available within the first hours of hospitalization.

Material and methods. This retrospective observational study included 1,144 patients diagnosed with non-ST-elevation myocardial infarction or unstable angina, hospitalized at a regional vascular center between 2019 and 2021. Invasive coronary angiography was performed in 900 patients, with ACO diagnosed in 64 (7.1%). To identify predictors of ACO, we analyzed 149 clinical, anamnestic, laboratory, and instrumental parameters, available during the initial hours of hospitalization. After selecting predictors, predictive models were developed using machine learning techniques: multivariate logistic regression (MLR), random forest, stochastic gradient boosting (XGBoost), and categorical boosting (CatBoost). Model performance was evaluated based on ROC-AUC, sensitivity (Sens), specificity (Spec), Precision-Recall AUC (PR-AUC), Brier-score, positive predictive value (PPV), negative predictive value (NPV), F1-score, and Accuracy.

Results. Independent predictors of ACO included newly diagnosed or new left ventricular regional wall motion abnormalities, erythrocyte sedimentation rate >15.5 mm/h, high-density lipoprotein cholesterol <0.9 mmol/L, creatine phosphokinase MB >82 U/L, and signs of ongoing myocardial ischaemia at admission: refractory or recurrent chest pain and/or dyspnea, combined with at least one other established criterion of very high-risk NSTE-ACS according to current clinical guidelines. The CatBoost-based model demonstrated the best prognostic performance (ROC-AUC = 0.841). A highly interpretable prognostic model based on multivariable logistic regression, using the identified risk factors, had performance metrics comparable to the CatBoost model.

Conclusion. The combination of clinical, laboratory, and echocardiographic parameters enables effective prediction of ACO in the NSTE-ACS cohort. The developed machine learning models demonstrate diagnostic accuracy and can form the basis for modifying early risk stratification algorithms and optimizing invasive treatment strategies.

Aim. To evaluate the role of clinical parameters, pre-test probability (PTP) of coronary artery disease (CAD), and its modifying factors in predicting the risk of cardiovascular events in a modern Russian cohort of patients with suspected CAD using conventional statistical methods and artificial intelligence (AI).

Material and methods. The prospective observational study included 210 patients (115 men (54,8%), aged 60,1±10,1 years). The PTP of coronary artery disease (CAD) was assessed. Glucose, lipid profile, and creatinine levels were determined. Electrocardiography, carotid and femoral artery ultrasound was performed. The prospective follow-up period was 21 [19-25] months. Cardiovascular endpoints (CVEs) included cardiovascular death, acute coronary syndrome, and myocardial revascularization. Statistical analysis was performed using Statistica for Windows, version 16.0 (StatSoft, USA). Prognostic models were built using the Python programming language (version 3.x) and the Scikit-learn, Pandas, and NumPy machine learning libraries.

Results. The PTP of CAD was 17% [11-26%]. Prognostic data were obtained for all patients (100%), with CVE recorded in 51 of them (24,3%). By a univariate analysis, a higher risk of CVE was associated with CAD PTP, carotid and femoral atherosclerosis; by a multivariate Cox regression — with CAD PTP and femoral atherosclerosis (sensitivity — 63%, specificity — 64%, accuracy — 64%, AUC — 0,68, p<0,001). AI methods revealed that CAD PTP and triglyceride (TG) levels were independent predictors of CVEs (sensitivity — 61%, specificity — 66,7%, accuracy — 69%, AUC — 0,70, p<0,001). The TG level had a direct impact on the prognosis in the cohort with CAD PTP of 16-23% as follows: the TG ≥1,7 mmol/L was a marker of unfavorable prognosis. The personalized prognosis was calculated using the equation "CVE probability=1/(1+exp(-z))", where z=-1,5674+(0,0592×CAP PTP, %)+(0,1871×TG); a value >0,53 (53%) indicated an unfavorable prognosis.

Conclusion. Using AI technologies, we established that CAD PTP serves as the most accurate indicator for the primary cardiovascular risk stratification and personalized referral for specific diagnostics in suspected CAD. In addition to CAD PTP, the detection of femoral atherosclerosis is associated with an increased risk of CVE. TG levels are of additional importance for decision making in the cohort with a CAD PTP of 16-23%.

Aim. To compare diagnostic effectiveness and interpretability of two following approaches to electrocardiogram analysis in amyloid cardiomyopathy (ACM): a deep learning model (1DResNet) and an algorithm for vectorcardiogram (VCG) loop planarity assessment..

Material and methods. The study was conducted on a dataset including electrocardiogram recordings from patients with verified ACM (n=99) and a control group (n=2673). Augmentation and undersampling were used to balance the classes. Two following methods were employed: a convolutional neural network (CNN) with the ResNet architecture and the Grad-CAM decision interpretation. QRS and ST-T segment planarity assessment (R2) was performed with  random forest classification.

Results. The 1D ResNet model demonstrated stable pathology detection with sensitivity and specificity of 75%. The Random Forest classifier, based on planarity characteristics, demonstrated higher performance as follows: specificity — 85%, predictive accuracy (precision) — 79%, sensitivity — 75%. Activation map analysis (Grad-CAM) revealed that the T-wave was the most significant signal area for the neural network in amyloid cardiomyopathy (93,75% contribution). This is consistent with the planarity method, where features of the repolarization segment also made the main contribution (57%) to the decision.

Conclusion. Both methods confirm that key diagnostic abnormalities in amyloid cardiomyopathy are located in the repolarization phase (T-wave). The planarity based approach provides a physical interpretation (spatial heterogeneity), while Grad-CAM visualizes the diagnostic logic of the neural network, which coincides with pathophysiological assumptions.

Aim. To describe the experience of developing a guideline-based web tool by a regional vascular center (RVC) team to re-check patients with atrial fibrillation and streamline referral to an arrhythmologist consultation regarding catheter ablation, created without professional programmers using neural networks at the development stage, as well as to assess changes in RVC process metrics in comparable pre-/post-implementation periods.

Material and methods. The web tool was developed in November 2025 and implemented in December 2025. A domestic AI system GigaChat (https://giga.chat/) was used for AI-assisted development (interface and code generation/debugging), while the clinical logic (input fields and routing rules) was predefined by clinicians based on the analysis of Russian and international clinical guidelines. Pilot implementation covered the following cities of Perm Krai: Berezniki, Solikamsk, Chaikovsky, Kungur, Lysva, and Kudymkar. Physician participation was voluntary, while access was provided via a shared link. No patient identifiers were entered or stored (name, date of birth, address, insurance number, etc.). The output was transferred by the physician into the Perm Krai regional health information system at their discretion. Following comparable periods were analyzed: January–February 2025 vs January–February 2026. Count outcomes were compared using an exact method for Poisson rates with equal exposure; rate ratios (RR) with 95% confidence intervals (CI) are reported.

Results. Arrhythmologist consultations at the RVC increased from 102 to 134 (RR=1,31; 95% CI 1,02-1,70; p=0,043). Radiofrequency ablation increased from 34 to 46 (RR=1,35; 95% CI 0,87-2,11; p=0,219), and cryoablation from 30 to 38 (RR=1,27; 95% CI 0,78-2,04; p=0,396). Total ablation procedures increased from 64 to 84 (RR=1,31; 95% CI 0,95-1,82; p=0,118). Total repeat procedures increased from 20 to 35 (RR=1,75; 95% CI 1,01-3,03; p=0,058).

Conclusion. Using neural networks at the development stage enabled clinicians to rapidly create and deploy a guideline-based web tool for organizational routing of atrial fibrillation patients to catheter ablation consultation. The tool was intentionally not positioned as a full clinical decision support system and does not replace specialist decision-making; its role is guideline-based re-checking and standardized output entry into the Perm Krai regional health information system.

Aim. To develop and describe in detail a method for multiscale computational assessment of restenosis risk after carotid endarterectomy, integrating imaging data, blood flow modeling taking into account formed element rheology, vascular biomechanics, neurogenic regulation, and genomic and epigenomic profiling. Restenosis after carotid endarterectomy remains an unsolved problem in vascular surgery, developing in a significant proportion of patients undergoing surgery. Existing prediction methods lack accuracy due to the inability to integrate all pathogenesis factors (hemodynamic, biomechanical, neurohumoral, and molecular genetic).

Material and methods. This work presents a following step-by-step methodology: acquisition of data from multislice computed tomography angiography and 4D phase-contrast magnetic resonance imaging; construction of threedimensional geometric models; simulation of erythrocyte, leukocyte, and platelet motion using dissipative particle dynamics; hyperelastic finite element modeling of the vascular wall with fluid-structure interaction analysis; modeling of neurogenic regulation incorporating a baroreceptor feedback loop; tensor analysis of hemodynamic fields; whole-genome sequencing; DNA methylation analysis; microRNA profiling; and the development and training of a hybrid neural network architecture.

Results. The proposed methodology enables the integration of nine independent data modalities into a unified predictive system. High accuracy in restenosis prediction was achieved (AUC 0,995). The methodology can be reproduced at any research center equipped with the appropriate hardware and software.

Conclusion. The proposed methodology represents the first fully integrated approach to multiscale restenosis prediction, combining imaging, hemodynamic, biomechanical, neurophysiological, genomic, and epigenomic data.

Aim. To assess the frequency of use and level of trust in physical examination (PE) methods, as well as the degree of artificial intelligence (AI) integration into the practice of specialists managing arrhythmias.

Material and methods. An anonymous online survey of 143 respondents (cardiology and surgical physicians and residents) was conducted. The questionnaire included questions on professional and demographic data, frequency of use and trust in nine PE methods, and the use of AI tools. Statistical analysis was performed in RStudio using Fisher’s exact test, Pearson’s χ2 test, Spearman/Pearson correlation analysis, and PAM (Partitioning Around Medoids) (Gover distance) clustering.

Results. The analysis included 127 fully completed questionnaires (88% physicians, 12% residents). The most frequently used PE methods were cardiac/lung auscultation and blood pressure measurement, while the least frequently used were cardiac percussion and chest palpation. Frequency of use correlated with the type of work as follows: cardiologists used a wider range of techniques, while invasive specialists (radiologists, cardiovascular surgeons) used only a limited number of FEs. The main reasons for not using FEs were the availability of paraclinical methods (41,7%) and time constraints (44,9%). Only 40% of respondents use AI in their work, primarily for information retrieval and writing. Cluster analysis revealed three following specialist phenotypes: "conservative" (broad physical examination, no AI), "technology-oriented" (selective physical examination, active AI), and "surgical" (narrowly selective physical examination).

Conclusion. A following significant gap in diagnostic approaches was identified: a decline in physical skills among young and highly specialized physicians is combined with a predominantly "technical" use of AI. The obtained data substantiate the need to integrate a hypothesis-driven approach to physical examination and AI training into medical education programs to maintain clinical competence.

Aim. To assess performance and methodological quality of machine learning (ML) models for predicting outcomes in patients with heart failure with preserved (HFpEF) and reduced (HFrEF) ejection fraction.

Material and methods. A systematic search of original publications in PubMed and eLibrary (2015-2025) was conducted. Twenty studies were included in the review. Methodological quality was assessed using PROBAST.

Results. The performance of the ML models was comparable for HFpEF and HFrEF (median AUC, 0,812 in both groups, p=0,48). The analysis showed that 70% of the studies were at high risk of bias, primarily due to the lack of external validation. The clustering models demonstrated clinically meaningful stratification, identifying phenotypes with significantly increased risk (e.g., hazard ratio 2,99 (95% confidence interval 2,41-3,7, p<0,001)).

Conclusion. ML models demonstrate moderate-to-high performance. However, the lack of external validation limits their readiness for clinical application. Standardization of methodology for future studies and prospective validation are needed.

This article analyzes current data on the potential of artificial intelligence (AI) technologies in diagnostics, cardiovascular risk stratification, and personalized treatment of patients with hypertension (HTN). An analytical review of publications on the use of AI and machine learning methods in HTN was conducted. A literature search was conducted in PubMed, Scopus, Web of Science, ScienceDirect, and Google Scholar. The review included publications In Russian and English from 2020-2026 using the following keywords: artificial intelligence, machine learning, hypertension, risk stratification, prediction, and diagnosis. AI algorithms demonstrate high potential in the diagnosis and monitoring of hypertension by analyzing data from electronic health records, 24-hour blood pressure monitoring, electrocardiography, retinal imaging, and wearable devices. AI use opens new prospects for hypertension management, including early diagnosis, more accurate risk stratification, and personalized therapy. AI is not a substitute for clinical judgment, but it is becoming a powerful tool for decision support. Further clinical trials, external validation of algorithms, and standardization of health data are necessary for the widespread adoption of these technologies.

The introduction of artificial intelligence (AI) technologies pushes the boundaries in cardiac surgery, but safe clinical practice requires systematization of data on its capabilities and limitations. The study aim is to systematize current data on the use of AI in cardiac surgery and identify promising areas for its clinical implementation. This systematic review of data from PubMed, Scopus, the Cochrane Library, Google Scholar, and Web of Science for the period 2000-2025 in accordance with the PRISMA criteria analyzed studies on AI use at all stages of cardiac surgery. Based on the analysis of 179 studies, machine learning models demonstrate higher sensitivity compared to traditional diagnostic methods and risk scores in predicting postoperative outcomes and complications. AI-based robotic systems and computer vision can improve the precision of surgical interventions, and the use of AI for postoperative monitoring can improve patient outcomes and rehabilitation. The main barriers to scaling these technologies remain data insufficiency, ethical considerations, and the difficulty of integrating them into clinical practice. Thus, AI has the potential to improve the quality of cardiac surgical care. However, realizing this potential requires validating algorithms, eliminating systemic errors, and developing transparent ethical and legal standards.

This literature review systematizes and critically analyzes artificial intelligence (AI) tools used in experimental and translational cardiology, assessing the prospects for their practical use and the corresponding limitations. Currently, AI tools are most actively used to automate the search, systematization, and analysis of information, as well as to perform mathematical calculations during the experimental planning stage. Research on AI algorithms for automated analysis of text and graphical datasets based on machine learning is aimed at reducing the time required for preclinical studies to accelerate the translation of cardioprotective, anti-atherosclerotic, endothelial-protective, and anti-calcification pharmacological interventions into clinical practice. In particular, AI algorithms are capable of automatically identifying morphological structures and performing their morphometric assessment during the analysis of biological tissues and cell cultures. AI tools have high potential for revealing hidden and complex patterns in tabular data from omics studies, enabling the identification of intermolecular interactions and the objective reconstruction of the development of typical pathological processes. The active use of AI code generators to create specialized computer programs eliminates the need for interdisciplinary collaboration in the automation of experimental data processing.

Cardiovascular and neurological diseases are leading causes of morbidity and mortality worldwide, and their close pathophysiological relationship, known as the brain-heart axis, requires a comprehensive multidisciplinary approach to diagnosis. This review analyzes and summarizes research conducted primarily over the past five years on the integration of artificial intelligence technologies into the diagnostics of this group of diseases. The review reveals that artificial intelligence, particularly deep learning models, demonstrates transformative potential in the analysis of electrocardiograms, neuroimaging, and multimodal clinical data, providing significant improvements in accuracy and early detection of pathologies. Key advances include the ability of artificial intelligence algorithms to identify hidden disease markers inaccessible to human perception and to predict the risk of conditions such as atrial fibrillation and ischemic stroke. However, the widespread clinical implementation of artificial intelligence faces significant challenges, including black box, systemic bias in training data, poor model generalization, and a severe lack of evidence from large-scale prospective clinical trials. The paper concludes that realizing the potential of artificial intelligence for personalized predictive medicine in neurocardiology is only possible by overcoming existing technical, ethical, and regulatory barriers through interdisciplinary collaboration.

Aim. To review the literature on the effectiveness of artificial intelligence (AI) technologies for screening, diagnosis, and monitoring of cardiovascular diseases.

Material and methods. We searched publications in PubMed, Web of Science, Scopus, CyberLeninka, eLibrary, and Google Scholar. The search strategy included the following keywords: "circulatory system diseases", "cardiovascular diseases", "artificial intelligence", "machine learning", "deep learning", "patient monitoring", "remote monitoring". Inclusion of original studies from 2015 to 2025 was based on independent author assessment.

Results. Of 594 publications, 8 studies meeting the inclusion criteria were included in the final analysis.

Conclusion. AI is a tool that is transforming modern methods of monitoring, diagnosing, and predicting cardiovascular outcomes. AI-based solutions demonstrate high diagnostic and prognostic efficacy, often exceeding traditional clinical scores, and form the basis of intelligent decision support systems for physicians.

Despite the fact that coronary angiography is the gold standard for diagnosing coronary stenosis, assessment of their hemodynamic significance in most cases remains operator-dependent. Currently, methods that analyze intracoronary physiology parameters are recommended for the objectification of measurements. With the rapid development of artificial intelligence (AI) technologies, new solutions for non-invasive assessment of hemodynamic parameters are emerging, some of which have already been validated in large studies. The aim of this review is to analyze and systematize published data on the use of AI methods in the non-invasive assessment of coronary artery hemodynamic parameters. This review utilized publications indexed in PubMed, Google Scholar, Web of Science, Cyberleninka, and E-Library. The search covered a 10-year period, beginning in 2016. The review is based on summarized data from the most relevant clinical studies and systematic reviews. A literature review concluded that the results of using AI technologies to assess coronary artery hemodynamic parameters are comparable to those of classical invasive techniques. However, further development and improvement of this approach remain a pressing research challenge.

A trend of the last decade has been the use of artificial intelligence (AI) technologies aimed at developing preventive medicine by optimizing the use of healthcare resources, including reducing the burden on medical personnel and improving the accuracy of medical data analysis. This review highlights AI-based tools, current trends, and prospects for their application in interpreting electrocardiogram patterns to identify changes often missed by standard routine analysis. Accumulated data from several studies indicate the prognostic value of AI-based models trained on sinus rhythm electrocardiograms in the diagnosis of paroxysmal atrial fibrillation, hypertrophic cardiomyopathy, and heart failure.

Aim. To evaluate the literature on digital auscultation, comparing the effectiveness of human auditory system and machine analysis.

Material and methods. A search of publications was conducted in PubMed, eLibrary, and Google Scholar using the following keywords: "digital auscultation", "digital phonocardiography", "sound analysis during auscultation", "neural network", "artificial intelligence", "machine learning".

Results. Of the 64 publications screened, 15 studies meeting the inclusion criteria were included in the final analysis.

Conclusion. The effectiveness of digital technologies and artificial intelligence for the automatic diagnosis of heart murmurs has been confirmed. Further development is focused on improving the analysis of phonocardiographic signals, advancing telemedicine, and creating universal platforms for population screening.

Acute ischemic stroke remains a cause of high mortality and disability. Existing treatments are limited by a small therapeutic window. Fingolimod has demonstrated neuroprotective potential, but its use is limited by cardiovascular risks (bradycardia) due to S1PR3 activation. Siponimod is a selective S1PR1/S1PR5 modulator without activity on S1PR3. Analysis of the mechanisms of action, clinical data, and pharmacoeconomic modeling indicate that siponimod may be more effective and safer than fingolimod in ischemic stroke. Artificial intelligence modeling of a clinical trial shows that the use of siponimod in 1000 patients could prevent disability in 510 people and save over RUB1,6 billion. The ICER is 11765 RUB/QALY. Thus, siponimod is a promising candidate for repurposing with a favorable cardiac profile.

Aim. To systematize data on the use of machine and deep learning algorithms for diagnosis, prognosis, and personalization of therapy in cardiac patients.

Material and methods. A systematic search of PubMed, Scopus, Web of Science, Google Scholar, and eLIBRARY.RU was conducted for the period 2015-2026. Inclusion criteria included original studies, clinical focus, use of AI algorithms, and model validation. The selection process was described in accordance with PRISMA 2009. Of the 187 identified publications, 41 were included in the final analysis after duplicate removal and screening.

Results. Five priority areas were identified: automated electrocardiogram analysis, echocardiographic image interpretation, computed tomography and magnetic resonance imaging data processing, cardiovascular risk assessment, and personalized treatment approaches. Neural network algorithms demonstrate high accuracy, comparable to expert judgment. Fundamental electrocardiography models (DeepECG-SSL) achieve an AUC of 0,990 on internal tests and 0,981-0,983 on external data. Multimodal approaches integrating retinal and cardiovascular data show an AUC of 0,97. The first randomized trial of large-scale language models in cardiology demonstrated a reduction in clinically significant errors from 24,3% to 13,1% (p=0,033).

Conclusion. Artificial intelligence is becoming a key tool in cardiac diagnostics and prognosis. Russian researchers are making a significant contribution to this field. Unresolved issues remain regarding model interpretability and the need for external validation.

The rapid integration of generative neural networks into the practice of medical research presents an unprecedented challenge to the system of academic certification. The ability to automatically generate coherent, scientifically plausible text calls into question the integrity of dissertation work and the validity of academic degrees as markers of genuine research competency. This paper analyzes the epistemological risks posed by "synthetic" dissertations, including the fabrication of data, the erosion of methodological transparency, and the degradation of established scientific schools. It examines contemporary methods for detecting AI-generated text — ranging from stylometric analysis to software-based detectors — and discusses their inherent limitations. As a systemic response to this crisis, a comprehensive set of measures is proposed: the introduction of mandatory in-person writing of the dissertation’s final section in a controlled environment with video surveillance; the tightening of quantitative criteria for admission to the dissertation defense (including H-index requirements, a multifold increase in the number of required publications, and a minimum length of professional service); and a fundamentally new model for the periodic recertification of academic degrees, similar to professional accreditation, which includes the possibility of degree revocation in cases of insufficient publication activity. The implementation of these measures would create a multi-layered system of protection for the body of medical science, preventing the infiltration of "synthetic" researchers and preserving the value of the academic degree as an authentic testament to scientific contribution.

The article provides an overview of the problem of using generative artificial intelligence tools when writing scientific articles. The article considers the movement of an article created by an AI algorithm within the editorial office of a scientific peer-reviewed journal and the measures to counteract violations of academic ethics that the editorial board can take. The analysis of modern systems for determining the generated text and a new tool for the authors’ work, which can help the editorial staff verify the manuscript under consideration, is carried out. The article concludes that it is necessary to review the approaches of journal editors in preparing the text of a scientific manuscript for publication.

Artificial intelligence (AI) in cardiology today is not so much a tool as a system capable of outperforming humans in data analysis, diagnostics, medical recommendations, and even in completing exam assignments. This raises the question of how future doctors should be trained, given that much of the current medical knowledge and skills can already be performed by machines? It has become evident that there is a need to rethink the very foundations of cardiology education using innovative educational technologies. This review is dedicated to these issues. Literature data suggest that it is necessary to shift the emphasis from rote memorization of facts to fostering critical clinical thinking, teaching logical thinking techniques, the capacity to make decisions in conditions of uncertainty, and the ability to collaborate with technology while maintaining human qualities such as empathy and spirituality. The review analyzes the changing role of AI in education and its impact on teaching and testing knowledge. It also considers the task of countering various negative effects of AI. The authors believe that within the ongoing digitalization the ability to empathize with and take into account the cultural and religious background of patients will remain the areas where the doctor is indispensable. Due to the obvious emergence of new competencies in the cardiologist, a paradigm shift in medical education is necessary in the near future.

This paper analyzes the prospects for using machine learning to assess cardiovascular risk in patients with type 2 diabetes. It is shown that modern algorithms, particularly neural networks, demonstrate high efficiency in risk prediction. However, despite the high accuracy of the models, their implementation is limited by several issues, such as the lack of external validation on independent samples and the difficulty of integrating them into clinical decision support systems. The authors emphasize the need for standardized study design and transparency in algorithm performance to implement machine learning in widespread clinical practice.

Neural interfaces are emerging as a promising area of development in modern medical technologies. Brain-computer interface (BCI) technologies and neuromodulation systems enable the direct recording and interpretation of neural activity, providing novel opportunities for diagnostics, rehabilitation, and restoration of lost functions. Evidence data notes that BCI use can significantly expand the potential of neurorehabilitation, communication, and motor function restoration in patients with severe motor impairments. At the same time, the effectiveness and safety of their clinical implementation largely depend on the level of technological maturity of the devices, the quality of clinical data, as well as regulatory frameworks and adherence to ethical principles.

This article examines the main types of neural interfaces and key areas of their medical application, including communication restoration in severe paralysis, neurorehabilitation after stroke, and motor function restoration in cases of nervous system damage. Particular attention is paid to clinical evidence analysis for these technologies, as well as an assessment of the socioeconomic, legal, and ethical aspects of their implementation in the healthcare system.

Artificial intelligence models show promising results in identifying new features of coronary artery occlusion. The study aim was to analyze and systematize literature on the use of artificial intelligence in the diagnosis and risk stratification of occlusion myocardial infarction.

Heart failure is a global healthcare problem: despite the availability of effective treatment regimens, only approximately 15% of patients receive optimal, evidence-based therapy. A key reason is the complexity of decision-making in time-constrained outpatient settings, which creates a demand for artificial intelligence (AI) tools. This article analyzes current AI developments in three areas of heart failure care: diagnosis, therapy optimization, and outcome prediction. International studies confirm the high potential of these algorithms, but most are still in the clinical validation stage. In Russia, there is a scientific foundation, but large-scale prospective studies, regulatory mechanisms, and the integration of these solutions into health information systems are lacking. Systemic barriers to implementation are identified and following steps to overcome them are proposed: developing and launching Russian analogues of international studies, advancing explainable AI, creating a national platform for testing algorithms, and training personnel. Implementing these areas will allow Russia not only to fill the existing gap but also to take a leading position in cardiac care digitalization.

Aim. To study AI’s current perceptions of its capabilities in cardiology and promising development directions, and to assess AI’s perceptions of potential limitations of its use.

Material and methods. A survey of two large Russian-language language models was conducted using 12 structured prompts divided into four thematic blocks: current capabilities, limitations, development prospects, and comparative analysis. Each prompt was sent three times to increase the reliability of the results. The models’ responses were analyzed and compared.

Results. Both models demonstrated a common vision of the future of AI in cardiology as a physician’s "digital partner", but their approaches to self-assessment differ significantly. GigaChat demonstrates a self-assessment of its technical and functional capabilities and limitations as a language model. Alice AI primarily describes the general opportunities and systemic barriers to AI implementation in medicine based on external literature.

Conclusion. The era of AI in cardiology has already arrived and will continue to evolve. Progress in the use of AI in cardiology depends primarily on training physicians to work in the new digital ecosystem and resolving legal issues. The ultimate goal of AI is to create a tandem between humans and artificial intelligence, where technology reduces routine work, enables rapid analysis of large volumes of data, and expands physician capabilities, ensuring increased accessibility and quality of personalized medical care for each patient.