POSSIBILITIES OF ARTIFICIAL INTELLIGENCE IN HEART FAILURE DIAGNOSIS

Aim. To summarize existing approaches to the use of artificial intelligence in the diagnosis of heart failure, to characterize the algorithms and models employed, to describe the types of medical data used (ECG, echocardiography, EMR, CT/MRI, angiography, wearables), to evaluate model performance (accuracy, AUC, sensitivity/specificity), and to assess feasibility and prospects for clinical implementation – with particular attention to the situation and challenges in Kazakhstan. Materials and methods . Systematic searches of PubMed , Scopus , Web of Science , IEEE Xplore and Google Scholar (2015 – 2025) identified peer-reviewed English and Russian studies on AI applications for heart failure diagnosis; two reviewers independently screened articles, extracted data and assessed quality, and results from 60 eligible studies were synthesized narratively with quantitative pooling where appropriate. Results and discussion. Across 60 eligible studies (2015 – 2025), AI applied to ECG, echocardiography, EMRs, imaging and wearable data demonstrated diagnostic accuracy typically between 85-95% (AUCs up to 0.97); ECGbased algorithms reliably detected HFrEF, AI-assisted echocardiography improved segmentation and reduced operator dependence, multimodal models enhanced prediction of therapy response (including CRT), while implementation in Kazakhstan remains nascent due to infrastructure and data-access limitations. Conclusion. Artificial intelligence is a promising direction in heart-failure diagnostics that can enhance the accuracy, timeliness and personalization of clinical decisions. For large-scale clinical adoption – especially in Kazakhstan – prospective validation, standardized protocols, local representative datasets, robust digital infrastructure and workforce training are required.