DETECTION OF ATRIAL FIBRILLATION BASED ON ECG - TESTING THE EFFECTIVENESS OF A SIMPLE AI ALGORITHM IN IDENTIFYING ARRHYTHMIAS ON ECG RECORDINGS

Atrial fibrillation (AF) is the most common cardiac arrhythmia and a growing cardiovascular epidemic characterized by uncoordinated atrial activation. It is a significant risk factor for ischemic stroke, heart failure, and cognitive decline. Despite its rising prevalence, driven by aging populations and lifestyle factors, early detection remains a diagnostic challenge due to the often asymptomatic and paroxysmal nature of the condition. Traditional screening methods, relying on standard electrocardiograms (ECG) and manual interpretation, are resource-intensive and limited in their ability to provide continuous monitoring. The rapid development of digital health technologies has introduced artificial intelligence (AI) as a pivotal tool for automating arrhythmia detection. This review assesses the effectiveness of "simple" machine learning algorithms, such as Support Vector Machines (SVM), Random Forests (RF), and k-Nearest Neighbors (k-NN), in comparison to complex deep learning architectures like Convolutional Neural Networks (CNN). The analysis indicates that simple AI models demonstrate high diagnostic accuracy (up to 99.1%), comparable to deep learning models, while requiring significantly less computational power and offering greater interpretability. These features make them highly suitable for deployment in battery-operated consumer devices, such as smartwatches and handheld ECG recorders. The integration of these efficient algorithms into telemedicine infrastructures supports a hybrid care model, facilitating large-scale screening and early diagnosis, which is essential for mitigating the burden of AF-related complications. Methodology: This article is a review. The data presented is based on a comprehensive analysis of peer-reviewed scientific articles, systematic reviews, and clinical reports published in reputable medical journals. The review focuses on studies evaluating the performance of AI algorithms using publicly available physiological signal databases, such as the MIT-BIH Atrial Fibrillation Database and PhysioNet Challenge datasets, as well as data collected from wearable mobile devices. The literature selected for this overview covers the period from 2010 to 2025, including foundational research and the most recent developments in mobile health technologies, approved medical devices, and algorithmic efficiency. The review critically compares the utility, accuracy, and computational requirements of various machine learning approaches in the context of remote patient monitoring and clinical diagnostics.