A Novel Ischemia Detector via ECG Signals Utilizing Support Vector Machine and Extreme Learning Machine Techniques

This study aims to develop a new robust method to distinguish ischemic ST-T episodes from normal parts in order to detect ischemia. The method operates by applying a hybrid technique based on wavelet transform and low-pass filter in order to remove redundant noise and baseline wandering, calculation of a representative average for each episode called Mean-Signal, and then extraction of efficient feature vectors from QRS complex and P and T-wave sections of each representative signal and at last differentiating ischemic episodes from normal ones using two classifiers. Using European ST-T database for train and test, the algorithm was evaluated by Extreme learning machine and support vector machine classifiers. Obtained sensitivity and positive predictive value for SVM method were respectively 91% and 85.6%. These values for extreme learning machine are respectively 96.8% and 90.65%.

کلیدواژه ها:

Ischemia ، ELM ، SVM ، Entropy ، DWT ، Wavelet energy ، Notch