Heart disease diagnosis using discrete wavelet transform and artificial neural networks

Abstract : This thesis shows an approach for ECG signal processing based on
Artificial Neural Networks (ANN) and transform domains (Discrete Wavelet Transform(DWT) and Fourier Transform(FT)), in order to perform automatic analysis using personal computers. The Neural Networks (NN) are ntroduced to solve different pattern recognition problems associated with ECG analysis. A Multi-Layer Perceptron (MLP) NN is used in the present work with Back Propagation (BP) algorithm to train the proposed network. The operation of a classification system based on DWT and FT was analyzed for diagnosing a set of real patterns (QRS interval of normal and MI disease) that were taken from many patients of MAC-1200 12 channel ECG system from Al-Iskandariya General Hospital and Al-Kadmia Teaching Hospital. Then an off-line method was used for the extraction of ECG signals from ECG images papers by using special image processing techniques. MATLAB software package version 6.5 was used to implement the proposed algorithms. The work in this thesis describes the analysis of ECG signals by using DWT coefficients and FT (magnitude and phase). The resulted coefficients are fed to the NN classifier for classification of the extracted feature vectors. The heart and its electrical aspect that include ECG wave generation, sensors, leads of the body and all types of noise are shown in the present work. Also the Ischaemia and Myocardial Infarction (MI) are dealt with in details. The obtained accuracy for the WT-NN is 90.