Page 132 - YORAM RUDY BOOK FINAL

P. 132
of the inverse solution. This resulted in a marked improvement of accuracy, as quantified by the

relative error of reconstruction (relative to measured epicardial potentials). Oster also developed a
method to incorporate spatial characteristics of the potentials in the regularization procedure
(IEEE Transactions on Biomedical Engineering 1997;44:188) 276 , called “regional regularization”. In a
model study, this approach was shown to improve epicardial potential reconstruction by up to 25%.

In reconstruction of repolarization and its spatial dispersion, Ghanem (Circulation
2001;104:1306) 277 incorporated spatial dependence through the use of Tikhonov second-order
regularization. This higher order regularization scheme constraints the curvature in space of the

epicardial potentials rather than their magnitude. With this method, T waves were reconstructed
with greater accuracy and preserved amplitudes. Repolarization properties and dispersion,
introduced by regional myocardial warming and cooling, were faithfully reconstructed.

Validation of ECGI Reconstructions in Humans

Evaluating Accuracy of Locating Initiation Sites, Based on Known Pacing Sites Locations.

Ventricular Pacing:

1. GMRes with quadratic interpolation (Ghosh, Annals of Biomed Eng 2005;33:1187) 273
– Patient 1: RV apical pacing site was reconstructed within 10 mm of pacing lead location

as determined from CT (based on reconstructed epicardial potentials; from isochrones it
was 12 mm, less accurate as expected). Patient 2: apical RV pacing – 2 mm error,
LV pacing – 7 mm error. Patient 3: LV pacing – 4 mm error. Patient 4: RV pacing – 2 mm error;
LV pacing – 3 mm error. All reconstructions captured the general spread of activation from

the pacing site.

2. Method of Fundamental Solutions (MFS) (Wang, Annals of Biomed Eng 2006;34:1272) 274 .
RV Pacing: 8 mm error with BEM and 5 mm with MFS. Simultaneous RV and LV pacing

during bi-V pacing in CRT: RV pacing site - 5.2 mm error; LV pacing site - 7.4 mm error.

3. L-1 Norm (Ghosh, Annals of Biomedical Engineering 2009; 37:902) 272 . Bi-V pacing from
epicardial leads in pediatric patients with congenital heart disease. Patient 1 (6 years old;

congenital heart block): RV pacing site – 2 mm error; LV pacing site – 6 mm error (individual
pacing from each lead). Patient 2 (5 years old, corrected transposition of the great arteries
and heart block): Pulmonary ventricle pacing site – 4 mm error; systemic ventricle pacing
site – 6 mm error. Patient 3 (17 year old, single double inlet LV post Fontan and congenital AV

block): Posterior lead – 3 mm, anterior lead – 5 mm. Over 5 patients with dual lead pacing;
the accuracy of L-1 norm was 3.8 ± 1.5 mm and of L-2 norm with zero order Tikhonov
10 ± 2.8 mm.