Classification of Abnormal Cardiac Rhythm from Brief Single-Lead ECG Recordings using Embeddings from Transformer Encoder Models.

Electrocardiograms (ECGs) are universally recognized diagnostic tools in cardiology, providing clinicians with graphical representations of the voltage differences generated during the cardiac cycle. This non-invasive procedure has become essential in diagnosing a wide range of cardiac conditions, including arrhythmias, myocardial infarctions, heart failure, and structural anomalies such as cardiomyopathy and hypertrophy. Traditionally, physicians acquire the skill of interpreting ECGs through extensive training, enabling them to identify specific patterns indicative of various disease states.