Automated identification of atrial fibrillation from a short single-lead electrocardiogram using the heart rate variability characteristics of the cardiac cycle

Atrial fibrillation (AF) is a commonly observed cardiac arrhythmia in clinical contexts, accounting for roughly one-third of hospitalizations for cardiac rhythm disorders. The present epidemiological evidence has substantiated that Atrial Fibrillation (AF) is a prevalent worldwide epidemic that has deleterious effects on both enduring morbidity and mortality. The manual interpretation of an electrocardiogram (ECG) for diagnostic purposes is a demanding and time-consuming process that requires a significant level of expertise. Furthermore, the issue is characterized by inconsistencies among various observers and even within a single observer. The identification of Atrial Fibrillation (AF) presents difficulty owing to its frequent asymptomatic nature and the necessity of conducting multiple diagnostic examinations to verify its existence. The lack of adequate documentation or identification of atrial fibrillation (AF) and its associated comorbidities is a cause for concern. Atrial fibrillation (AF) has the capacity to give rise to significant medical complications, such as stroke and heart failure, among others. The proposed study aims to create a machine learning algorithm that has low computational requirements and can accurately identify Atrial fibrillation in its early stages using short single lead electrocardiogram signals in near real-time. The framework under consideration has been primarily developed with a focus on the Heart Rate Variability (HRV) characteristics of the human cardiac cycle. The framework underwent training and testing on the MIT-BIH Atrial Fibrillation and CinC 2017 Database and demonstrated a sensitivity of 95.30% and specificity of 94.71% through the implementation of a ten-fold cross-validation approach. The framework was subjected to experimentation using medical samples obtained from a local laboratory, and the results indicate an accuracy rate of approximately 91.5%. The proposed method could help doctors find atrial fibrillation early on during routine ECG screenings.