The Evolution of Artificial Intelligence in Cardiology: From Precise Arrhythmia Detection to Prediction of Cardiovascular Events and Mortality Risk - A Systematic Review (2018–2026)

Background: Cardiovascular disease is the leading cause of death globally, with cardiac arrhythmia being one of the main causes of adverse clinical events. In recent years, psychophysiological medicine has undergone a major transition; however, the first step in this process is the development of Artificial Intelligence (AI) for predicting cardiovascular risk and establishing a framework for prognostic modeling. Aim: This systematic review aimed to evaluate how AI has been applied in Electrocardiography (ECG) in the period from 2018 to 2026. More specifically, we sought to determine the accuracy of AI in diagnosing arrhythmias and the emerging use of AI for the prediction of major cardiovascular events and death. Methods: A systematic search of the MEDLINE (PubMed), Embase and Cochrane Library databases for original studies published from 2018 to 2026 was conducted. The focus was on deep learning-based algorithms that utilized 12-lead ECGs or signals derived from wearables. The risk of bias and quality of the studies was assessed using the QUADAS-2 tool. Results: There were 30 initial studies included in the overall analysis. Most models were convolutional neural network (CNN) based and demonstrated excellent diagnostic accuracy as demonstrated by sensitivity and specificity commonly reported >90–95%. AI's ability to detect paroxysmal atrial fibrillation while in a normal sinus rhythm and for superior triage of occlusive myocardial infarction (STEMI equivalent) versus standard clinical criteria are important findings. Additionally, the establishment of "AI-ECG age" as a digital biomarker may enable long-term prediction of cardiovascular mortality and heart failure risk. Though there are substantial differences in methodological quality and limited standardized external validation among diverse populations are prominent obstacles. Conclusion: This makes AI a revolutionary tool within cardiology where it can deliver results of accuracy equal to or greater than specialists on selective tasks. The transition from diagnostic to prognostic systems creates tremendous promise for the development of personalized medicine. Yet, in order to achieve appropriate integration into clinical practice, more standardized reporting, as well as rigorous validation to ensure patient safety and diminish the "black box" characteristics of deep learning models, must be undertaken within real-world, multi-national cohort studies.