A Comparative Analysis for Heartbeat Signal Classification based on Metaheuristic Approaches

Published: 13-01-2021| Version 1 | DOI: 10.17632/n7zbdnn8fj.1
Juan Carlos Carrillo-Alarcón,
Ignacio Algredo-Badillo,
Héctor Rodríguez-Rángel,
Mariana Lobato-Báez,
Mariano Vargas-Santiago,
Luis Alberto Morales-Rosales


In the literature, we found metaheuristic approaches for heartbeat classification consisting of parameter optimization. These metaheuristic approaches have been applied in two contexts: employing a binary classification and using multiclass classification with a specific number of classes. There exists a significant reduction in the performance of classifying heartbeat when the datasets are unbalanced. In this paper, we present an evaluation of two metaheuristic optimization approaches based on a multiclass heartbeat classification in the presence of unbalanced classes. We focus on classifying heartbeat classes ranging from 2 to 8 for intra-patient cases and obtain competitive results versus the state-of-the-art works, obtaining up to 99\% performance since we look for broad detection of heartbeat types such as non-ectopic beats, supraventricular ectopic beats, and ventricular ectopic beats with higher specificity, sensitivity, precision, and accuracy.


Steps to reproduce

Code for classifying heartbeat classes using metaheuristic approaches. The scripts in MATLAB uses the database MIT-BIH Arrhythmia Database (https://physionet.org/content/mitdb/1.0.0/). Scripts: MITDB1N_DatToMat: To change to MATLAB format. MITDB2N_Filtrado: Preprocessing, to filter the signals. MITDB3N_Normalizacion: Preprocesing, to normalize the signals. MITDB4N_Segmentacion_HL: Preprocessing, to segment the filtered signals by HL filtering. MITDB4N_Segmentacion_Wav: Preprocessing, to segment the filtered signals by Wavelet filtering. MITDB5N_Clusters: Preprocessing, to cluster the majority classes. MITDB6N_FeatureExtraction: To extract features of heartbeats. MITDB7N_DE: To optimize using differential evolution approach. MITDB7N_DE: To optimize using particle swarm optimization approach.