Ventricular surface activation time imaging from clinical data: a quantitative comparison of two inverse approaches

Ventricular activation time maps for both the endocardium and epicardium were estimated from MRI and body-surface ECG mapping data which had been acquired under clinical conditions. The associated nonlinear inverse ill-posed problem was solved by two different approaches. The first approach (method A) was based on a linearization technique while the second (method B) employed an optimization routine for nonlinear ill-posed problems. Both methods utilized a second-order Tikhonov regularization in order to stabilize the inverse solution. The linearization technique could be demonstrated to perform better, particularly with regard to clinical implementations and applications. In two patients suffering from Wolff-Parkinson-White syndrome, the activation time maps were estimated and quantitatively compared.