Document Type : Original Article
Authors
1
Electrical Engineering group, Dept. of Engineering, Islamic Azad University of Mashhad, Mashhad, Iran
2
Department of Electrical Engineering, Faculty of Engineering, Islamic Azad University Mashhad Branch, Mashhad, Iran
3
Cardiology group, Dept. of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran
4
Cardiovascular Surgery group, Dept. of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran
Abstract
Atrial fibrillation (AF) is one of the most common arrhythmias, a lack of timely diagnosis of which can result in stroke or even mortality. To date, various techniques have been used to recognize this type of arrhythmia. However, deep learning has been one of the most commonly used methods in this area, which has attracted the attention of researchers in recent years. Meanwhile, the long short-term memory (LSTM) has had an efficient performance among deep neural networks used in this area owing to its continuous analysis feature for analyzing time-series data such as ECG. Therefore, the present study applied an LSTM structure with the least number of layers at each level has been used, in order to reduce network learning time, thereby, increasing the detection speed diagnosing the desired problem. Also the propose design, were discussed and comparison with two highly applied deep learning of BILSTM and CNN, and the results were indicative of higher accuracy(86%) and detection speed(2 and 19 times faster) of LSTM compared to the other two structures. In addition, in this research by calculating the correlation coefficient for the 3 parameters of age,sex and heart rate the effectiveness of each parameter in identifying the patient was examined.
Keywords
Main Subjects
)