ATRIAL FIBRILLATION, WEARABLE TECHNOLOGIES AND ARTIFICIAL INTELLIGENCE: CURRENT EVIDENCE, CLINICAL UTILITY AND LIMITATIONS

Abstract

Background: Atrial fibrillation is the most common sustained cardiac arrhythmia in adults and represents an important clinical and public health problem because of its association with stroke, mortality, disability, and healthcare burden. Its paroxysmal and frequently asymptomatic course limits the effectiveness of conventional diagnostic strategies based on occasional electrocardiographic recordings. Aim: This narrative review aims to summarise current evidence regarding the use of wearable technologies and artificial intelligence for the detection of atrial fibrillation, with particular emphasis on diagnostic accuracy, clinical utility, and principal limitations. Methods: A structured literature search was performed using PubMed, Scopus, Web of Science, and Google Scholar. Studies published between 2021 and 2026 were primarily considered. Search terms included “atrial fibrillation,” “wearable technologies,” “smartwatch,” “electrocardiography,” “photoplethysmography,” “artificial intelligence,” “machine learning,” and “diagnostics.” Priority was given to original studies, systematic reviews, meta-analyses, and clinically relevant review articles concerning wearable devices, artificial intelligence, and atrial fibrillation detection. Earlier landmark studies were included when they were necessary to explain key evidence discussed in recent reviews and meta-analyses. Results: Wearable devices such as smartwatches, fitness bands, and portable ECG systems allow longer heart rhythm monitoring in daily life and may help detect paroxysmal and clinically silent atrial fibrillation. ECG-based devices are more consistent with current diagnostic standards, whereas photoplethysmography-based systems are more accessible and can provide passive rhythm monitoring. Artificial intelligence, especially machine learning and deep learning, can support signal interpretation, arrhythmia recognition, and risk prediction. However, diagnostic performance reported in controlled studies may not always translate directly into real-world clinical practice. Conclusions: Wearable technologies and artificial intelligence offer a promising way to support atrial fibrillation screening and monitoring. They may improve early detection, especially in asymptomatic individuals, but should not replace standard diagnostic methods. Photoplethysmography-based alerts should be regarded as screening signals, while a final diagnosis of atrial fibrillation should still be confirmed by ECG in an appropriate clinical context. More large-scale studies are needed before these tools can be widely implemented in routine practice.