Section: New Results

Inverse problem

We tested our method using synthetic data generated with a highly realistic forward model. Propagating action potentials were generated using a monodomain reaction-diffusion model with a Ten Tusscher 2006 membrane model. An anisotropic human heart model at 0.2-mm resolution was used for this purpose. Torso potentials were then computed from the simulated transmembrane currents using a finite-difference torso model at 1-mm resolution with intracavitary blood, anisotropic myocardium, lungs, and an anisotropic skeletal muscle layer. We simulated 20 cases: 5 single stimuli, 1 dual stimulus and 14 re-entry simulations. From the simulated torso potentials a 200-channel body surface map recording was extracted and used to test the inverse methods. Inverse solutions in terms of epicardial potentials were computed both with MFS and with our .new optimal control approach. With our algorithms, we were able to construct the electrical potential on the heart surface with a very good accuracy in terms of correlation coefficient. Thus, we could accurately reconstruct the activation pattern.