Atrial Fibrillation Detection by the Combination of Recurrence Complex Network and Convolution Neural Network

Abstract

In this paper, R wave peak interval independent atrial fibrillation detection algorithm is proposed based on the analysis of the synchronization feature of the electrocardiogram signal by a deep neural network. Firstly, the synchronization feature of each heartbeat of the electrocardiogram signal is constructed by a Recurrence Complex Network. Then, a convolution neural network is used to detect atrial fibrillation by analyzing the eigenvalues of the Recurrence Complex Network. Finally, a voting algorithm is developed to improve the performance of the beat-wise atrial fibrillation detection. The MIT-BIH atrial fibrillation database is used to evaluate the performance of the proposed method. Experimental results show that the sensitivity, specificity, and accuracy of the algorithm can achieve 94.28%, 94.91%, and 94.59%, respectively. Remarkably, the proposed method was more effective than the traditional algorithms to the problem of individual variation in the atrial fibrillation detection.

An Automatic System for Real-Time Identifying Atrial Fibrillation by Using a Lightweight Convolutional Neural Network

“…It can be seen that the proposed lightweight CNN tends to outperform those algorithms without considerations of cardiac rhythms. Wei et al [38] achieved an accuracy of 94.59% for AF detection by using a CNN based classifier with initial features of recurrence complex network, while a longer training time of 9.65 h is needed for this 6-layer CNN. Meanwhile, Xu et al [39] and Xia et al [28] reported two CNN based detectors with more than 10 layers with an accuracy of 84.5% and 98.63%, respectively.…”

“…However, the training process can be carried out off-line. In this work, the proposed lightweight CNNs by using two cardiac rhythm features as input can be trained in less than 10 minutes to converge with 2,000 back propagation iteration, and the training time is also less than others previous works [28], [38], [40], as shown in Table 4. Two of the important reasons that we used less time for training possibly include there two initially extracted representative rhythm features and a lightweight CNN model, which both could speed up the deep learning of hidden features and obtain a faster training of our proposed model.…”

“…As for the testing process of our lightweight CNNs, several experiments were performed on the recorded 24 h clinic data to obtain the time cost, and it was revealed that the testing process is about 0.004 s for one sample of 10 s data and 0.25 s for 24 h data, which shows a practical capability in real-time AF detection. Most recently, Wei et al [38] used a CNN based method for AF detection to process one sample (one heartbeat) in 0.1186 s, and Andersen et al [40] reported that a 24 h single-lead data can be classified in 0.92 s. Both of these two studies also reported that their method could be used to detect AF in real-time processing. It should be noted that the low level features extraction in previous works and the rhythms features extraction in this work would be the most time-consuming process (35 s for 24 h data in our work), especially for the earlier works [28], [39] by using two dimensional wavelet transform or recurrence complex network.…”

HADLN: Hybrid Attention-Based Deep Learning Network for Automated Arrhythmia Classification

“…The effective analysis of ECG signals is helpful to detect many heart diseases such as atrial fibrillation (AF), myocardial infarction (MI), and heart failure (HF) ( Turakhia, 2018 ). In an AF waveform, the P wave is replaced by many inconsistent fibrillatory waves, and the RR interval is irregular, which is easily mixed with other diseases ( Wei et al, 2017 ). In the early stage, the research work of ECG classification was generally implemented by using manual feature extraction method.…”

An ANN Ensemble Based ECG Signal Classification Approach for Accurate Arrhythmia Detection

“…In [8], different kinds of arrhythmia have been classified by employing an ANN based technique which exploits morphological and dynamic features. The classification of different categories of cardiac arrhythmias has also attracted the use of advanced ANN based algorithms such as Deep Neural Networks [9 -10], Deep Convolutional Neural Networks (CNN) [11], hybridization of CNN and Recurrence Complex Networks [12], etc. The use of a hybrid approach combining learning approaches like the deep/ensemble learning, and evolutionary optimization techniques to enhance the accuracy of arrhythmia detection has been recently reported by [13].…”