Automated Signal Pattern Detection in ECG During Human Ventricular Arrhythmias

Ventricular arrhythmias seriously affects cardiac function. Of these arrhythmias, Ventricular fibrillation is considered as a lethal cardiac condition. Recent studies have reported that ventricular arrhythmias are not completely random and may exhibit regional spatio-temporal organizations. These organizations could be indicative of reoccurring signal patterns and might be embedded within the surface electrocardiograms (ECGs) during ventricular arrhythmias. In this work, we aim to identify such reoccurring ECG signal patterns during ventricular arrhythmias. The detection of such signal patterns and their distribution could be of help in sub-classifying the affected population for better targeted diagnosis and treatment. Our analysis on 14 ECG segments (on average 3.24 minutes per segment) obtained from the MIT-BIH ventricular arrhythmia database identified three reoccurring signal patterns. A wavelet based technique was developed for automating the pattern identification process using ECGs. The proposed method achieved automated detection accuracies of 73.3%, 75.0% and 86.6% for the proposed signal patterns.