Genetic algorithm with Gaussian function for optimal P-wave morphology in electrocardiography for atrial fibrillation patients

Atrial fibrillation (AF), which is a rapid-irregular heartbeat and shows abnormal heart rhythm of the supraventricular tachycardia class, has proved to increase the risks of heart failure, dementia, and stroke. To detection AF, P-wave morphology in electrocardiography (ECG) is suggested to be a strong indicator. To make computerized detection possible, most approaches decompose the ensemble of signals into a finite set of features and establish the relation between symptoms and values of features. Therefore, the disease can be asserted solely by the values of the decomposed features. For early diagnosis of AF, this study develops a hybrid Taguchi-genetic algorithm (HTGA) that facilitates Gaussian decomposition in ECG signals, because P-wave morphology can be well approximated by a family of Gaussian functions. The HTGA optimizes features with minimized performance index of the normalized root mean square error. With accurate decomposition in characterizing parameter values of P-wave morphology, the performance of disease classification improves by using appropriate feature set. Our experiments indicate that the proposed HTGA with Gaussian function obtains a better fit to the actual P-wave compared to the conventional nonlinear least-squares approaches.