THE TRANSFORMATION OF CARDIOVASCULAR SCREENING IN ATHLETES: THE MULTIMODAL ROLE OF ARTIFICIAL INTELLIGENCE IN DIFFERENTIATING CARDIAC PATHOLOGIES FROM PHYSIOLOGICAL ADAPTATION

Purpose: The primary objective of this review is to evaluate the efficacy, clinical applications, and current limitations of Artificial Intelligence (AI) and Machine Learning (ML) in diagnosing cardiovascular diseases (CVD) among competitive athletes. Specifically, this study addresses the critical diagnostic challenge of differentiating benign physiological adaptations known as "athlete's heart" from potentially lethal pathologies, including cardiomyopathies and channelopathies, to prevent sudden cardiac death. Materials and Methods: A systematic literature search was conducted across PubMed, Scopus, and Web of Science databases covering the period from 2000 to 2025. The review identified and synthesized 48 key studies utilizing AI algorithms—specifically deep learning applied to electrocardiography (AI-ECG) and automated imaging analysis (Echocardiography, CMR)—for the detection of Hypertrophic Cardiomyopathy (HCM), Arrhythmogenic Right Ventricular Cardiomyopathy (ARVC), valvular anomalies, and inherited channelopathies. Diagnostic performance metrics were analyzed to compare AI methodologies against standard clinical criteria. Results: Deep learning models applied to ECG demonstrate superior sensitivity (>90%) in detecting occult cardiomyopathies compared to traditional methods , while AI-enhanced imaging significantly improves the reproducibility of tissue characterization. AI algorithms, such as those analyzing phonocardiograms, show efficacy comparable to echocardiography in detecting valvular heart disease. Conclusions: AI represents a paradigm shift in sports cardiology, offering potential for scalable and cost-effective screening. However, widespread clinical implementation is currently hindered by the "black box" nature of algorithms and the scarcity of large, athlete-specific training datasets. Future deployment requires explainable AI models validated on diverse athletic cohorts.