Clustering of source locations based on the transform matrix for ECG simulations

Abstract

In this work the number of dipole sources required for accurate ECG simulation was estimated. Locations for dipole sources were clustered so that sources in each cluster were transformed to ECG by the same matrix. The clustering was applied to an infinite homogeneous conductor model and an inhomogeneous human thorax model both for the standard 12-lead ECG simulation. In the latter model 19, 116, 326, and 678 clusters were formed from 7046 locations in cardiac muscles keeping the relative mean square errors less than 10, 4, 2, and 1%, respectively, from simulations by transforming the sources individually.