Novel genetic ensembles of classifiers applied to myocardium dysfunction recognition based on ECG signals

This article presents an innovative genetic ensembles of classifiers applied to classification of cardiac disorders (17 classes) based on electrocardiography (ECG) signal analysis.

From a social point of view, it is extremely important to prevent heart diseases, which are the most common cause of death worldwide. According to statistical data, 50 million people are at risk for cardiac diseases worldwide.

This research collected 744 fragments of ECG signals from the MIT-BIH Arrhythmia database for one lead, MLII, from 29 patients. Novel methodology that consisted of the analysis of longer (10-s) fragments of the ECG signal was used (an average of 13 times less classifications). To enhance the characteristic features of the ECG signal, the power spectral density was estimated (using Welchs method and a discrete Fourier transform). In research designed two genetic ensembles of classifiers optimized: by classes and by sets, based on: SVM classifier, 10-fold cross-validation method, ensemble learning, layered learning, genetic selection of features (frequency components), genetic optimization of classifiers parameters and novel genetic training (selection of experts votes) used to combining classifiers.

The best genetic ensemble of classifiers optimized by sets, obtained a classification sensitivity of 17 heart disorders (classes) at a level of 91.40% (64 errors per 744 classifications, accuracy = 98.99%, specificity = 99.46%, time for classification of one sample = 0.0186 [s]). Against the background of the current scientific literature, these results represent some of the best results obtained.