Atrial fibrillation detection based on multi-feature extraction and convolutional neural network for processing ECG signals. (April 2021)

• Record Type: Journal Article
• Authors: Chen, Xianjie; Cheng, Zhaoyun; Wang, Sheng; Lu, Guoqing; Xv, Gaojun; Liu, Qianjin; Zhu, Xiliang
• Is Part Of: Computer methods and programs in biomedicine. Volume 202(2021)
• Publication Date: 2021-04
• Subjects: Atrial fibrillation detection -- Atrial activity signal -- Convolutional neural network -- Accuracy -- Multiple feature extraction; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2021.106009
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: An atrial fibrillation detection algorithm based on multi-feature extraction and CNN is proposed. The accuracy, specificity, sensitivity, and true positive rate are used as evaluation criteria. Compared with cluster analysis, one-to-one rule and support vector machine detection. The accuracy of our algorithm this paper is as 98.92%, which is 23.25% higher than the comparison algorithm. Abstract: Background and objective: The incidence of atrial fibrillation is increasing annually. We develop an automatic detection system, which is of great significance for the early detection and treatment of atrial fibrillation. This can lead to the reduction of the incidence of critical illnesses and mortality. Methods: We propose an atrial fibrillation detection algorithm based on multi-feature extraction and convolutional neural network of atrial activity via electrocardiograph signals, and compare its detection based on cluster analysis, one-versus-one rule and support vector machine, using accuracy, specificity, sensitivity and true positive rate as evaluation criteria. Results: The atrial fibrillation detection algorithm proposed in this paper has an accuracy rate of 98.92%, a specificity of 97.04%, a sensitivity of 97.19%, and a true positive rate of 96.47%. The average accuracy of the algorithms we compared is 80.26%, and the accuracy of our algorithm is 23.25% higher than this average pertaining to the other algorithms. Conclusion: We implemented an atrial fibrillation detection algorithm that meets the requirements of high accuracy, robustness and generalization ability. It has important clinical and social significance for early detection of atrial fibrillation, improvement of patient treatment plans and improvement of medical diagnosis.