Ischemic Episode Detection in an ECG Waveform Using Discrete Cosine Transform and Artificial Neural Network

Abstract

A new method that uses discrete cosine transform (DCT) and artificial neural network (ANN) is presented for the detection of myocardial ischemic(MI) episodes in ambulatory ECG (AECG) monitoring based on the ST-T segment changes. First, ST-T segments are extracted based on the detection of the R-peaks in the AECG waveform. DCT is then used to represent the extracted ST-T segments. A subset of the DCT coefficients is used as a feature vector to represent the ST-T segments. Finally, a three-layered feedforward ANN trained with backpropagation algorithm is used to classify the ST-T segments as normal or abnormal (potential MI episodes). In computer simulations, the classification rates as high as 82% were achieved. The results show that the DCT based artificial neural network (ANN) is a viable approach to detect ischemic episodes using ST-T segments of AECG waveform.