ECG source location clustering based on position vectors and forward transfer matrices

Heart model segmentation methods concerning ECG source-to-measurement forward transfer modeling are discussed. The k-means clustering technique was adopted to classify all discrete points forming a heart model with respect to their position vectors or source-to-measurement transfer matrices. The clusters were formed in heart models of end-systolic and end-diastolic cardiac phases. The minimum number of clusters for different lead systems, cardiac phases and in volume conductor models determined for least square error approximation of nonclustered (original) transfer model are tabulated. These numbers suggested that the number current dipole sources could be reduced to less than 10% of that of the source locations. Some of the heart models segmented by the resulting clusters are presented at the end of this article.