Dynamic registration of cardiac US and CT data using Fourier descriptors and Dynamic Time Warping

Cardiac Resynchronization Therapy (CRT) can be optimized by the fusion of anatomical, functional and electrical information in a unified framework in order to identify the most effective pacing sites. The aim of this work is to perform a registration between dynamic CT and ultrasound (US) images (2D speckle tracking mode). The proposed registration approach is based on a contour to surface scheme, decomposed in four steps: (1) the temporal synchronization of data; (2) the segmentation of the left ventricle endocardial surface from the 3D+t CT data; (3) the segmentation and tracking of the myocardium from the 2D+t US data; (4) the registration of US contour with the CT surface. The originality of the method relies on the use of Fourier descriptors and Dynamic Time Warping (DTW) to handle different spatial and temporal resolutions as well as dissimilar cardiac rhythms between CT and US data. An evaluation on simulated data is described as well as results obtained on three patient databases.