Identification of high risk patients in cardiology by wavelet networks

Endangered patients with ventricular tachycardia (VT) could be identified by quantification of the low level high frequency content of their preprocessed ECG signals. For this purpose the authors developed a classification method which consists of two parts: (i) a small number of features is dimensional input patterns, preprocessed ECG signal, by wavelet transformation. A number of easily interpretable parameters in the time-frequency plane controls this feature extraction process. The calculated features are regarded as inputs to a simple artificial neural network (ANN) which is used as classifier. Because the learning phase of the ANN is expanded from training the weight coefficients to the time-frequency parameters of the input nodes, one gets optimally tuned features. The result of 96% correct classification for identifying VT patients by the described wavelet network approach confirms the efficiency of this method