Review of ECG detection and classification based on deep learning: Coherent taxonomy, motivation, open challenges and recommendations. (April 2022)

• Record Type: Journal Article
• Authors: Chen, Shan Wei; Wang, Shir Li; Qi, Xiu Zhi; Samuri, Suzani Mohamad; Yang, Can
• Is Part Of: Biomedical signal processing and control. Volume 74(2022)
• Publication Date: 2022-04
• Subjects: Electrocardiogram (ECG) -- Detection -- Classification -- Deep learning; 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.2022.103493
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract: An electrocardiogram (ECG) is one of the most promising approaches used for the detection and classification of cardiovascular diseases (CVDs) in recent years. This work reviewed ECG detection and classification that used deep learning algorithms in medical technology applications concerning research motivations, challenges and recommendations. Target retrieval was performed on four databases, namely, Science Direct, IEEE, Web of Science and PubMed, by using the following keywords: 'electrocardiogram', 'deep learning', 'deep neural network', 'convolution', 'detection', and 'classification'. A total of 97 papers were finalised. (1) Most of the papers (75 or 77.3% of the total) focused on designing and developing algorithms and software based on deep learning that can automatically detect and classify CVDs. (2) The second category consists of papers that focused on the design of smart wearable devices and hardware-based on deep learning and ECG. This category includes 12 papers, accounting for 12.4% of the total. (3) The third category comprises articles that used ECG as a biological signal recognition method. Monitoring and predicting CVDs and improving the speed and accuracy of prediction can be enriched by developing new methods or optimising and improving existing ones. Six papers belong to this category, accounting for 9.3% of the total. (4) The remaining four papers (accounting for 4.26% of the total) are reviews and surveys related to this field. Thus, the previous literature features technology realisation and improvement in detection and classification technologies for CVDs under the current technical condition. By contrast, the security and privacy protection of technologies receive less attention. Moreover, the existing literature has rarely focused on the topic of embedding an ECG detection system into intelligent wearable devices that detect and monitor CVDs.