ABNORMAL ECG (ELECTROCARDIOGRAM) PATTERN DETECTION USING SAX (SYMBOLIC AGGREGATE APPROXIMATION) ALGORITHM

Abstract

Electrocardiograph (ECG) is a record of electric signal obtained from heart activity. Such data is very helpful to recognize heart abnormalities quickly. ECG signals usually consist of periodical patterns that can be identified to make a diagnosis by a medical expert. However, the number of ECG recording data is another problem in diagnosing the condition of the patient's heart.

The paper aims to recognize abnormal ECG patterns from a number of ECG data records using SAX (Symbolic Aggregate approXimation) algorithm to convert complex ECG data into a simple set of symbols that have a certain periodic pattern. The symbols consist of sequences of character that can be processed using a pattern detector to recognize repetitive or non-repetitive patterns. Non-repetitive patterns can be categorized as anomalies (abnormal patterns) that may be caused by heart abnormalities.

The SAX and pattern detector algorithms are developed using MATLAB software. The ECG signal data in this study was taken from the MIT-BIH Arrhythmia Database https://physionet.org/physiobank/database/mitdb/ which provides access to recorded data on ECG signals in digital form.

The system is designed not to replace the role of a doctor or cardiologist, but to help medical experts in analyzing a large number of ECG signal data that is too time consuming to be read directly manually.

The test results show that the proposed method can recognize abnormal (non-periodic) patterns and normal (periodic) patterns that can be further analyzed by medical experts to make a diagnose.

Keywords: Anomalies, Arrhythmia, EKG, SAX