Soft Computing as a tool for Classification of Cardiovascular Abnormalities

Abstract

Classification of Electrocardiogram (ECG) for Cardio-Vascular Abnormalities (CVA) in the process of diagnosis is inevitable. In this paper, we propose a scheme to integrate Principal Component Analysis (PCA) with Neural Networks (NN) for classification of ECG Signals. A Neural Network (NN) with Back Propagation Algorithm is deployed as classifier. ECG samples consisting of Normal signals and three abnormal signals are taken from physionet arrhythmias database for our experiments. The PCA is used to minimize ECG signals into weighted sum of basic components that are statistically mutual independent. Thus, PCA is used for dimensionality reduction of data. Here a comparison of performance of Neural Network (NN) and Principal Component Analysis (PCA) with Neural Network (NN) are investigated. Principal Component Analysis (PCA) eliminates the least considerable data values, hence helps in improving the performance in classification of ECG signals. The results obtained suggest that Principal Component Analysis (PCA) with Neural Network (NN) performance is faster and better than Neural Network (NN) Classifier alone.