CAPABILITIES OF ARTIFICIAL INTELLIGENCE IN MODERN RHEUMATOLOGY. LITERATURE REVIEW

Abstract

Relevance: The application of artificial intelligence (AI) in clinical practice opens new opportunities to improve diagnostic and prognostic accuracy in rheumatology, a field characterized by high disease heterogeneity and the complexity of interpreting visual and clinical data Search strategy: An analysis was conducted on 58 publications selected from 467 relevant sources (2015–2025), identified through PubMed, Google Scholar, and CyberLeninka databases. Included were original studies and review articles focused on the use of machine learning and deep learning methods in rheumatology. Results: AI is increasingly being implemented to address tasks related to early diagnosis, disease course prediction, and personalized therapy selection in rheumatoid arthritis, spondyloarthritis, systemic lupus erythematosus, and osteoarthritis. Convolutional neural networks, multi-omics approaches, and adaptive prediction models were found to be the most effective, demonstrating high accuracy (with AUC values up to 0.85). Despite this potential, clinical integration remains limited due to small training datasets, the need for external validation, and insufficient model standardization. Conclusions: AI holds significant practical value in rheumatology. Its integration into healthcare systems requires regulatory frameworks, medical staff training, the development of digital infrastructure, and interdisciplinary collaboration.