An artificial neural network combining linear and nonlinear features from short heart rate variability segments achieved a global accuracy of 96.6% in classifying five cardiac rhythms.
Does an artificial neural network using linear and nonlinear features from HRV signals accurately classify cardiac rhythms?
An artificial neural network utilizing linear and nonlinear features from short heart rate variability segments (16 and 32 RR intervals) can effectively classify various cardiac arrhythmias.
Abstract Recent advances in artificial intelligence have growing importance, with applications in robotics, control systems, biometric recognition, among others. Intelligent systems for medical support represent another important area, where cardiac diagnosis is a valuable possibility. This work deals with an intelligent system capable of identifying and classifying cardiac rhythms from heart signals extracted from electrocardiograms (ECGs). Artificial neural networks are employed, and different configurations are investigated. A dynamical perspective is adopted, using linear and nonlinear features in both the time and frequency domains. Five cardiac rhythms are selected to test the system: normal sinus rhythm, atrial fibrillation, sinus bradycardia, premature ventricular contraction, and ventricular tachyarrhythmia. The neural network is verified, and clinical cases are subsequently analyzed. Results show satisfactory classification of patients’ heart rhythms, consistent with the diagnoses offered by cardiologists in their studies.
Silva et al. (Wed,) conducted a other in Cardiac arrhythmias. Artificial Neural Network (ANN) with linear and nonlinear features vs. ANN with only time-domain linear features was evaluated on Global classification accuracy. An artificial neural network combining linear and nonlinear features from short heart rate variability segments achieved a global accuracy of 96.6% in classifying five cardiac rhythms.