Wavelet-based features for characterizing ventricular arrhythmias in optimizing treatment options

Abstract

Ventricular arrhythmias arise from abnormal electrical activity of the lower chambers (ventricles) of the heart. Ventricular tachycardia (VT) and ventricular fibrillation (VF) are the two major subclasses of ventricular arrhythmias. While VT has treatment options that can be performed in catheterization labs, VF is a lethal cardiac arrhythmia, often when detected the patient receives an implantable defibrillator which restores the normal heart rhythm by the application of electric shocks whenever VF is detected. The classification of these two subclasses are important in making a decision on the therapy performed. As in the case of all real world process the boundary between VT and VF is ill defined which might lead to many of the patients experiencing arrhythmias in the overlap zone (that might be predominately VT) to receive shocks by the an implantable defibrillator. There may also be a small population of patients who could be treated with anti-arrhythmic drugs or catheterization procedure if they can be diagnosed to suffer from predominately VT after objectively analyzing their intracardiac electrogram data obtained from implantable defibrillator. The proposed work attempts to arrive at a quantifiable way to scale the ventricular arrhythmias into VT, VF, and the overlap zone arrhythmias as VT-VF candidates using features extracted from the wavelet analysis of surface electrograms. This might eventually lead to an objective way of analyzing arrhythmias in the overlap zone and computing their degree of affinity towards VT or VF. A database of 24 human ventricular arrhythmia tracings obtained from the MIT-BIH arrhythmia database was analyzed and wavelet-based features that demonstrated discrimination between the VT, VF, and VT-VF groups were extracted. An overall accuracy of 75% in classifying the ventricular arrhythmias into 3 groups was achieved.

Reducing False Arrhythmia Alarms Using Different Methods of Probability and Class Assignment in Random Forest Learning Methods

“…Hence, in analysing such signals, the algorithm cannot rely solely on heart rate. The methods already proposed for detection of ventricular tachycardia include flutter/fibrillation approaches like autocorrelation analysis [11], wavelet transformations [12,13], sample entropy [14], machine learning methods with features derived from signal morphology and analysis of power spectrum [15], time-frequency representation images [16], empirical mode decomposition [17], or using the zero crossing rate combined with base noise suppression with discrete cosine transform and beat-to-beat intervals [18].…”

Structured prediction for differentiating between normal rhythms, ventricular tachycardia, and ventricular fibrillation in the ECG

“…Additionally, analysis was performed using non-overlapping segments of ECG, as long as 10 seconds. Whilst some previous studies [5], [6], [7], [8], [9], [10], [11], [12], [13] have attempted to perform multiclass classification, it has been pointed out that the experimental procedures were not properly conducted using out of sample patients [3], [4], or other experimental errors exist such as preselecting easy to classify examples [8], [9] from VT and VF categories.…”

Support Vector Machine and Artificial Neural Network Implementation in Embedded Systems for Real Time Arrhythmias Detection

“…For this reason, a real-time cardiac monitoring systems must to include a module for the detection of these arrhythmias. To detect VT and VF in real-time, various algorithms have been developed, such as fuzzy similarity-base approximate entropy (Xie et al, 2011), Wavelet based features (Balasundaram et al, 2011), phase space reconstruction (Sáenz and Bustamante, 2009), feature selection with Support Vector Machine (SVM) (Alonso-Atienza et al, 2012) and Artificial Neural Network (ANN) (Valenza et al, 2008).…”