Discrimination of Ventricular Arrhythmias using Machine Learning models

Electrocardiogram (ECG) is a device used to detect and record the electric signals generated from different heart conditions. Analyzing these ECG signals helps in the recognition of various heart problems. One such problem is Ventricular Arrhythmia. This condition is one of the most fatal cardiac diseases which could result in the patient’s death. Ventricular Tachycardia (VT) and Ventricular Fibrmation (VF) are two types of Ventricular Arrhythmias, this work here depicts the use of ECG signals for recognizing Ventricular Arrhythmias (VT and VF) with the help of the ECG Feature Description model. A new feature set is created by using Singular Value Decomposition (SVD) and Harmonic Phase Distribution, to apprehend the morphological distinction of the ECG signal. Extraction of phase features is done with the help of the Fourier transform, Empirical Mode Decomposition (EMD), and Singular Value Decomposition (SVD). Further to that, Dynamic Time Warping is used for measuring Temporal Distinction between ECG signals. The primary fold (first) classification includes the discrimination of Shockable (VAs) and Non-Shockable Arrhythmias (Non-VAs) in 2s-segment, 5ssegment, and 8s-segment. ECG segments following that are the secondary fold (second) classification which distinguishes VT and VF rhythms. Accuracy levels got improved in the 5s window with the help of machine learning models compared to the existing methods.