Detection of life-threatening arrhythmias using temporal, spectral and wavelet features

Cardiac arrhythmia detection at the initial stage saves the patient from sudden death caused due to cardiac arrest. Arrhythmia can be predicted by detecting Ventricular Tachycardia and Ventricular Fibrillation. There are many techniques and methods for the detection of arrhythmia. The system proposes a highly efficient VF detector. It uses 18 parameters extracted from the ECG as input. These parameters can be broadly classified into 4 categories. They are temporal, spectral, complexity and wavelet features. After a Feature Selection technique, which sorts the parameters based on the rank score obtained, classification is done by both Artificial Neural Network and Support Vector Machine and their performances are evaluated. The results shown that Support Vector Machine along with Feature Selection is better than Artificial Neural Network.