A Comparative Evaluation of Atrial Fibrillation Detection Methods in Koreans Based on Optical Recordings Using a Smartphone

Abstract

This paper evaluated three methods of atrial fibrillation (AF) detection in Korean patients using 149 records of photoplethysmography signals from 148 participants: the k-nearest neighbor (kNN), neural network (NN), and support vector machine (SVM) methods. The 149 records are preprocessed to calculate the root-mean square of the successive differences in the R-R intervals and Shannon entropy which are validated from x-means and Massachusetts Institute of Technology and Beth Israel Hospital database for the features for AF detection. A smartphone camera was used to obtain photoplethysmography signals. Clinicians labeled 29 records by referring to the electrocardiogram signals. These labeled records were used as a ground truth set to evaluate the accuracy of each method. In the experiments, the kNN, NN, and SVM methods achieved 98.65%, 99.32%, and 97.98% accuracies, respectively.