Neural network approach to the inverse problem of electrocardiography

Truncated singular value decomposition (SVD), zero order Tikhonov regularization (TIK) and the generalized eigensystem (GES) approaches to the inverse problem of electrocardiography are eigenvector expansion techniques. Both SVD and TIK use the same expansion vectors, while GES utilizes different expansion vectors. For all of these methods the expansion coefficients are chosen using a least squares fit to the body surface potentials and then some “modification”. In this paper we explore the use of neural networks for choosing the expansion coefficients