Improved Inference of Heart Potentials using the Multipole-Equivalent Method


Background. Spectral characteristics of body-surface ECGs in patients with myocardial infarction differ in those at risk for ventricular tachycardia (VT) compared to those not at risk. Analysis of inferred potentials near infarct zones may improve risk assessment, but their utility is limited by high relative error (RE). To reduce REs, we estimated the multipole moments of cardiac sources, from which we calculated pericardial potentials. We expected smaller errors in indirect inference with the multipole-equivalent method (MEM) than via direct inference with the conventional boundary-element method (BEM) because MEM requires estimating unknowns, which are fewer in number, are orthogonal, have reduced correlation, and are ranked by spatial frequency.

Methods. We compared BEM and MEM in a 7-layer eccentric spheres model and in a realistic heart- and torso-model We measured an adult male torso at over 2000 positions, along with the locations of 190 electrodes used in body-surface mapping, to within 2 mm with an Immersion Personal Digitizer. We also estimated heart location, size, and orientation from ultrasonic images registered to the body-surface coordinates. Known pericardial potentials were taken from measurements over QRS with a 90-electrode sock in an adult male. Body-surface ECGs were calculated from these measurements using a BEM forward-problem solution in a 91-node heart and 1026-node torso model.

Results. For noise from I to 20 %, heart location and radius errors of ±1 cm, REs in the 7-layer eccentric- spheres model were 0.34 ± 0.18 for MEM and 0.66 ± 0.58 for BEM with a posteriori Tikhonov regularization. In the realistic heart-torso model REs were 0.61 ± 0.03 for MEM and 1.44 ± 1.62 for BEM.

Conclusions. Estimation of the multipole moments of cardiac sources with subsequent calculation of pericardial potentials yielded a reduction in RE by a factor of 2 compared to direct estimation of potentials in both eccentric-spheres and realistic-torso models. Significant reduction in REs of inferred potentials offers the prospect of better identification of risk for VT and other conditions.


Electrical and Computer Engineering


Poster session III

International Standard Serial Number (ISSN)


Document Type

Article - Journal

Document Version


File Type





© 2000 Elsevier, All rights reserved.

Publication Date

01 Jan 2000