Individualization of a vectorcardiographic model by a particle swarm optimization

Abstract

This paper presents the application of a bio-inspired method for optimizing a lifelike vectorcardiographic (VCG) model. During the model estimation, a Particle Swarm Optimization (PSO) seeks the optimal combination of all parameters that maximize the correlation coefficient (r) and minimize the Mean Squared Error (MSE) between the synthetic and directly measured VCG leads. The proposed method was tested on 52 different VCG records annotated as a healthy control (HC) from PTB database. 156 models were individualized without any previous analysis of the waves of the original records. The PSO method automatically provides very realistic models with a correlation coefficient r > 0.995 and MSE < 0.0005 mV2 for 152 of the 156 VCG signals.