Abstract
The growing number of wearable electrocardiogram (ECG) users today, combined with the surge of artificial intelligence (AI) and machine learning (ML) in medical signal-processing, has led to a new age of wearable-enabled monitoring for cardiac conditions. With the development of advanced processing methods, wearables offer the opportunity to monitor and predict the probability of various cardiac conditions, from cardiac ischemia to arrhythmias, by collecting personalized data from the comfort of a user's home. Although such technology has not yet entered the market, AI and ML research training specifically on wearable-based ECG data has grown significantly in the last decade. Despite this growing niche, there are few current articles reviewing the applications of these techniques in wearable ECG technology. To fill this gap, this article first primes the reader to the practical tools required to build models from ambulatory ECG, synthesizes the state of the field across major cardiac condition use-cases, and finally highlights recurring limitations in the current literature and outlines the need to improve reliability if this technology were to be widely utilized. As a result, we aim to help readers who otherwise may be unfamiliar with the specifics of these tools and their applications to form an interpretation of the current capabilities of AI/ML in wearable ECGs and identify key steps required for improvement based on the most current research.