Harmonic Analysis and Pattern Classification of Electrocardiograms for Heart Disease Diagnosis

Heart disease is a critical issue in improving people's health. Medical research and technology are being developed to obtain accurate diagnoses and treatments. This paper contributes to designing an automated diagnosis system to classify electrocardiogram (ECG) signals to detect cardiac diseases. With respect to other related works, it has the following distinctive characteristics: it is feasible to be implemented in real time, capable of detecting different heart pathologies, and effective in performance. The proposed system is based on Fourier series analysis, employing a dynamical state observer to instantaneously obtain salient features and patterns from the ECG harmonic content, whose information is classified through a K-nearest neighbor algorithm (KNN), named as the classifier, which determines the possible disease. The ECG signals used in this paper are obtained from the freely available PhysioNet databases, containing data to diagnose and classify healthy patients, arrhythmia cases, myocardial infarction, and heart failure. The proposed automated procedure is 93% effective in disease detection for the explored databases, highlighting its potential as a classification tool for ECG-based diagnosis.