A Novel Method for Detection of Atrial Fibrillation Based on Heart Rate Variability

Atrial Fibrillation (AF) detection is one of most important part of clinical testing. We propose a novel method to detect AF episodes based on heart rate variability feature of AF. In this method, scatter plot of the heart rate is found and histogram of the Y axis data of scatter plot is used for calculations. Depending upon amount of data present in each bin of histogram ECG signal is classified as Atrial Fibrillation (AF) or Normal Sinus Rhythm (NSR). Physionet 2017 challenge database, MIT-BIH AF database and MIT-BIH NSR database are used to validate the algorithm. Physionet Challenge contains 5787 ECG records of 30/60 s classified as AF or NSR, MIT-BIH AF database contains 25 full length ECG records and MIT-BIH NSR database contains 18 full length ECG records. Using the method, we got the accuracy of 97.23% for Physionet 2017 challenge database and 97.15% for MIT-BIH AF database. MIT-BIH NSR database didn't show any AF episode. This method can also be used for real time monitoring of ECG for AF detection.