R-Peak Detection of Beat Segmentation and Convolution Neural Network for Arrhythmia Classification

Erwin, Erwin (2022) R-Peak Detection of Beat Segmentation and Convolution Neural Network for Arrhythmia Classification. Journal of Engineering Science and Technology (JESTEC). ISSN 1823-4690 (In Press)

[thumbnail of Jestec R-Peak Detection.pdf]
Preview
Text
Jestec R-Peak Detection.pdf

Download (968kB) | Preview

Abstract

One of the abnormalities in cardiology is arrhythmia. The arrhythmia classification can use Electrocardiogram (ECG) signals. ECG has a large size because ECG should be recorded within a certain interval. The ECG arrhythmia can identify which beats have normal and abnormal patterns. The arrhythmia ECG data is provided by www.physionet.org which contains data from 47 patients in MIT-BIH that were recorded for 30 minutes. From the data, there are 650,000 signal points. The segmentation carried out in this study uses R-peak Detection. The segmentation results were 109,452 beats with 252 signals. In this study, R-peaks of ECG signals were combined with Convolution Neural Network (CNN) to detect and classify a normal and abnormal beat or beat that has arrhythmia. CNN is known as a robust classification method for data with large dimensions. The Accuracy result obtained from R-peak Convolution Neural Network (CNN) for arrhythmia classification on testing was 97%. The precisions for normal and abnormal beats are 99% and 91%. The sensitivities of normal and abnormal beats were 98% and 94%. These results indicate that the application of R-peak of ECG signals and CNN are excellent for arrhythmias detection on the ECG signals.

Item Type: Article
Subjects: Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation.
Divisions: 09-Faculty of Computer Science > 56201-Computer Systems (S1)
Depositing User: Dr Erwin Erwin
Date Deposited: 24 Feb 2022 06:06
Last Modified: 24 Feb 2022 06:06
URI: http://repository.unsri.ac.id/id/eprint/65294

Actions (login required)

View Item View Item