Paper
Abnormality Detection in Electrocardiograms by Time Series Alignment
- Authors:
- Bachir Boucheham
- Abstract
- Electrocardiogram (ECG) morphology deviation from normal beat is a sign of abnormal behavior and a significant information for cardiologist to depict cardiac diseases. Most existing methods for such a task use single beat classification with various tools. However, these approaches usually ignore the repetitive nature of the ECG. In this study, we adapt and apply a previously developed method for quasi-periodic time series comparison (SEA) to detect the morphology change in the ECG. The basic idea is to perform segment-wise comparisons of the ECG where one segment stands for the reference (normal) behavior and the other segment for the unknown behavior segment. Due to many difficulties, this is a very complex problem to solve, especially with regard to the phase shift and the number of periods in each segment problems. The new approach is applied on records from the Massachusetts Institute of Technology – Beth Israel Hospital (MIT-BIH) arrhythmia database. Results show the effectiveness of the proposed method in detecting the significant morphology changes of the ECG. The author believes that the method could also be useful for clustering and summarizing of ECG, among other applications.
- Keywords
- Electrocardiograms (ECG); Anomaly detection; Pattern recognition; Time series comparison; Shape Exchange Algorithm (SEA)
- StartPage
- 6
- EndPage
- 10
- Doi