Electrocardiogram Two-Dimensional Motifs: A Study Directed at Cardio Vascular Disease Classification

Aldosari, Hanadi, Coenen, Frans ORCID: 0000-0003-1026-6649, Lip, Gregory YH ORCID: 0000-0002-7566-1626 and Zheng, Yalin ORCID: 0000-0002-7873-0922
(2023) Electrocardiogram Two-Dimensional Motifs: A Study Directed at Cardio Vascular Disease Classification. In: Communications in Computer and Information Science. Springer Nature Switzerland, pp. 3-27. ISBN 9783031434709

[img] Text
ic3kBookAldsari_2023.pdf - Author Accepted Manuscript
Access to this file is embargoed until 16 September 2025.

Download (411kB)


A process is described, using the concept of 2D motifs and 2D discords, to build classification models to classify Cardiovascular Disease using Electrocardiogram (ECG) data as the primary input. The motivation is that existing techniques typically first transform ECG data into a 1D signal (waveform) format and then extract a small number of features from this format for classification purposes. It is argued here that this transformation results in missing data, and that the consequent feature selection means that only a small part of the original ECG data is utilised. The approach proposed in this paper works directly with the image format, no transformation takes place. Instead, motifs and discords are extracted from the raw data and used as features in a homogeneous feature vector representation. The reported evaluation demonstrates that more effective classification results than that which can be achieved using the waveform format. The proposed 2D motif and discord extraction mechanism is fully described. The proposed process was evaluated using three distinct ECG data sets. A best accuracy of 85% was obtained, compared with a best accuracy of 68.48% using a comparable 1D waveform approach.

Item Type: Book Section
Divisions: Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User: Symplectic Admin
Date Deposited: 19 Jun 2023 08:27
Last Modified: 22 Nov 2023 10:27
DOI: 10.1007/978-3-031-43471-6_1
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3171027