A Robust Detection Method of Atrial Fibrillation

Abstract

Atrial fibrillation (AF) is a common atrial arrhythmia occurring in clinical practice and can be diagnosed using electrocardiogram (ECG) signal. A novel method is proposed to detect normal, AF, non-AF related other abnormal heart rhythms and noisy recordings based on the combination of deep features and handcraft features. We used Computing in Cardiology Challenge 2017 database as training set and MIT-BIH atrial fibrillation database (AFDB) as test set. The proposed algorithm was achieved an accuracy of 96.3%, F1 of 95.5%, sensitivity of 88.7% and specificity of 99.6% in MIT-BIH AFDB, better than the method which only adopted deep features or handcraft features. Experimental results show that our method would be a good choice for the detection of the AF.