Linear and Nonlinear Optimization Methods for Noninvasive Functional Source Imaging

The reconstruction of cardiac electrical sources from body surface potential maps (BSPM) and from maps of the cardiac magnetic field MCG is of increasing scientific interest. The goal is the development of a noninvasive functional cardiac source imaging modality for clinical purposes. In a first step the epi- and endocardial activation time map during ventricular polarization is estimated from BSP and ECG data applying the critical point theorem. This result is then used as a starting vector for a successive optimization procedure. Several optimization strategies (e.g. BFGS) are investigated in order to increase the spatial and temporal resolution obtained. The ventricular AT pattern was estimated with a relative error of 0.1 from simulated ECG data after adding Gausian noise with a singal to noise ratio of 35dB.