Classification cardiac beats using arterial blood pressure signal based on discrete wavelet transform and deep convolutional neural network. (January 2022)

• Record Type: Journal Article
• Authors: Arvanaghi, Roghayyeh; Danishvar, Sebelan; Danishvar, Morad
• Is Part Of: Biomedical signal processing and control. Volume 71(2022)Part A
• Publication Date: 2022-01
• Subjects: Arterial Blood Pressure (ABP) -- Electrocardiography (ECG) -- Discrete Wavelet Transform (DWT) -- Scalogram -- AlexNet classification -- Convolutional Neural Network (CNN); 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.2021.103131
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract: Background and objective: Heartbeat type diagnosis in the early stage is the most crucial issue to survey patients and launch curating heart disorders. The traditional methods to diagnose heartbeat type depend on the clinician's judgment and experience. So, it companies some human mistakes in the diagnosis stage. To avoid human faults, some investigations are proposed to diagnose heartbeat types automatically. Because heart performance leads to some electrochemical (recorded as electrocardiography signal) and pressure (blood pressure waveform) signals in the whole of the body, so it sounds like its performance is assessable via mentioned signals. Methods: In this study, we have proposed a different signal to classify heartbeat types. We are using arterial blood pressure (ABP) signal instead of electrocardiogram signal to classify heartbeats in two groups of normal and abnormal types automatically. So, after denoising the ABP signal, its discrete wavelet transform (DWT) coefficient Scalograms are selected as the classifier input. In this study, the big challenge is signal type (ABP or electrocardiography (ECG)) to classify heartbeat. So, a deep convolutional neural network (CNN) is used to classify the ABP signal. Results: We have achieved 90.16% F1-score, 89.03% accuracy, 81.46% sensitivity, and 99.50% specificity in this study. Conclusions: It indicates the ABP signal has beneficial information about heart performance as efficient as the ECG signal.