Atrial Arrhythmia Detection Using LSTM and CNN

Document Type : Original Article

Authors

1 Assistant professor, Faculty of Electrical Engineering, K.N. Toosi University of Technology, Tehran. Iran

2 Faculty of electrical engineering, K.N. Toosi university of technology

Abstract

Atrial Fibrillation (AFib) arrhythmia is one of the most hazardous cardiac diseases. Accurate diagnosis of this arrhythmia can prevent many storks. In order to more accurate and early detection of this arrhythmia, automatic diagnosis is considered. In this paper, median filter and mean calculation of ECG signal were used for base line wander and muscle noise cancelation. The result of applying this method to ECG signal no.119 from MIT-BIH dataset, was 8.29 dB improvement in the SNR. Since variation in RR-interval of the ECG signal is one of the most common features of atrial fibrillation arrhythmia, RR-interval was considered as the input of algorithm instead of whole ECG signal samples. For R detection from ECG signal, Pan-Tompkins algorithm with adaptive threshold was used. Although in some other researches classic machine learning methods were utilized to detect AFib arrhythmia, deep learning method results in better accuracy, which is because of AFib detection dependency to sequential beats. Hence, in this article a deep learning network was designed using CNN and LSTM, also 10 RR-intervals were considered as the input data. The accuracy of inter-patient tests of this research on Challenge 2017 and MIT-BIH AFDB datasets were about 97% and 90%, respectively.

Keywords

Main Subjects

Advanced Signal Processing

Articles in Press, Accepted Manuscript
Available Online from 03 December 2023

Files

Share

How to cite

Statistics