Diagnosis of coronary artery occlusion by fitting polynomial curve with the ECG signal based on optimization techniques

The diagnosis of disease with patients having chest pain is a challenging issue for the general practitioners for further treatment without delay. When blood flow in a coronary artery is partially or totally restricted, it is called a coronary occlusion. It is essential to diagnose coronary artery blockage at the earliest possible stage to provide the appropriate therapy at the appropriate time. Various classification techniques have been used in many research studies to diagnose coronary artery disease. In this study, however, a polynomial curve fitting technique based on optimization strategies was used to diagnose coronary artery occlusion. The signals with the noise were removed using a discrete wavelet transform. This is followed by a hamming window function which shows the signals with 2000 samples. This was further applied to the polynomial fitting function to obtain polynomial coefficients of the given signal. The polynomial coefficients were obtained by choosing the best value of the polynomial order n using genetic and particle swarm optimization algorithm. Area, variance, kurtosis, root means, and form factor were computed using these polynomial coefficients. These features were further implemented in different classifiers such as Support Vector Machine, K-Nearest Neighbor, Levenberg-Marquardt Neural Network, Scaled Conjugate Gradient Neural Network, and Multilayer Perceptron Network. The K-Nearest Neighbor classifier provides the highest classification accuracy of other classifiers. The accuracy of the K-Nearest Neighbor classifier is 98.7 and 99.4 based on genetic algorithm and PSO technique respectively.

Automated summary

The study utilizes polynomial curve fitting technique and optimization algorithms to diagnose coronary artery occlusion. Various features are computed and implemented in different classifiers, and K-Nearest Neighbor classifier achieves the highest accuracy.