Segmentation-free MRI to CT 3D registration for Cardiac Resynchronization Therapy optimization.
Résumé
The purpose of this work is to include tissue and dynamic information from cardiac magnetic resonance (CMR) sequences in a previously proposed fusion framework aiming to optimize Cardiac Resynchronization Therapy (CRT). To do so, the 3D iconic registration between 3D+t cardiac MR and 3D+t cardiac computed tomography (CT) sequences is explored. Two rigid registration approaches have been evaluated: end-diastole (ED) images registration and dynamical time warping (DTW) based registration. DTW is used to align both sequences in time. They are tested on five patients that underwent for CRT. Quantitative evaluation has been performed using the dice score between ED delineations of left ventricle (LV). An average error of 4.71% (std 4.58%) is obtained for ED registration. For DTW registration, an average error of 2.68% (std 2.76%) is obtained using the normalized correlation curves of MR and CT sequences. These results demonstrate the feasibility to perform a 3D registration of MR and CT sequences without the need of temporal interpolation.