Photoplethysmograph based arrhythmia detection using morphological features. (March 2023)

• Record Type: Journal Article
• Authors: Neha, ; Sardana, H.K.; Kanawade, R.; Dogra, N.
• Is Part Of: Biomedical signal processing and control. Volume 81(2023)
• Publication Date: 2023-03
• Subjects: Photoplethysmography -- Signal processing -- Arrhythmia -- Machine learning algorithms; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.104422
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Automated detection of multiple types of arrhythmia i.e. PVC, ST, AFl, and Normal using PPG signals is proposed. Ground-truth generation for the data acquisition of PPG signals has been carried out using retrospective and prospective datasets. A new set of morphological features has been proposed for rule-based and statistical learning-based arrhythmias detection approaches. Comparison of statistical learning-based algorithms with rule-based approach on retrospective and prospective datasets. Abstract: Photoplethysmography (PPG) is a non-invasive optical technique that is used for the detection of cardiovascular diseases. The paroxysmal nature of arrhythmic events and the lack of timely recorded data emphasize the need to develop an automated method for the identification of arrhythmias. The literature shows the detection of a single type of arrhythmia using PPG. However, limited research has been carried out for the detection of multiple types of arrhythmia. In this research work, a new set of morphological features have been proposed for the automated detection of multiple arrhythmias using rule-based and statistical learning-based approaches. The proposed work has been implemented on the retrospective dataset and validated on the prospective dataset. The results show that the rule-based arrhythmia detection method is equipollent to the statistical learning approach with an accuracy of 98.43%/94.16% on the retrospective dataset and 94.16%/93% on the prospective dataset.