Improving the Dependability of the ECG Signal for Classification of Heart Diseases

Electrocardiogram (ECG) signals are widely used to classify a spectrum of human diseases, given the functional importance of the heart in the overall body activity. For automated classification, the degree of membership of an instance of ECG data to a particular disease can be measured using an indicator function. We use online classification and therefore this membership score is affected by fluctuations in the input data. A deterioration in the classification task is perceived when strong noise input signals are acquired or when disruption events qualitatively affects the input signal. The present work implements a method for detecting disruption events and noise in the ECG signal. The purpose of this method is to improve the quality and the reliability of the classification task for real measurements. The indicator function for disease classification using ECG data is based in the k-means cluster analysis method.