Meeting Banner
Abstract #1260

Automatic Registration of Multispectral MR Vessel Wall Images of the Carotid Artery

Ronald van 't Klooster1, Marius Staring1, Stefan Klein2, Robert Kwee3, Eline Kooi3, Johan Reiber1, Boudewijn Lelieveldt1, Rob van der Geest1

1Division of Image Processing, Department of Radiology, Leiden University Medical Center, Leiden, Netherlands; 2Biomedical Imaging Group Rotterdam, Depts. Of Radiology & Medical Informatics, Erasmus MC, Rotterdam, Netherlands; 3Department of Radiology, Cardiovascular Research Institute Maastricht, Maastricht University Medical Center, Maastricht, Netherlands

Patient movement during scanning can cause misalignments between multispectral MR images. Correctly aligned images of the carotid artery are needed to accurately classify plaque components inside the vessel wall. Manual alignment of these images is the current procedure but is time-consuming. Automatic image registration can be applied to correct for these misalignments. Various registration techniques were investigated and quantitatively validated with a gold standard. The results show that automatic image registration using a 3D B-spline transform can replace the manual alignment procedure reducing the amount of time needed analyzing carotid vessel wall images.

Abstract PDF Poster PDF

Keywords

accuracy acts addition adequately affine align aligned alignment allow amount analyzing applied applying around artery artifacts assessment atherosclerosis automated automatic available better bifurcation biomedical boundaries boundary cardiovascular carotid cause chosen circular classification close cohort coil columns comparing components consuming containing contour contours contrast correct descent described detailed develop diameter distance division dominant dutch equally evaluation expert fifty final five form foundation gold gradient grid ground heart informatics inside institute investigated linear lumen manual manually mask medical meshes metric misalignment misalignments molecular movement moving mutual neck needed nine none occurs optimal optimizer paired patient patients performance pixel plaque plot post prevent previous procedure processing publicly quantify quantitative quantitatively radiology randomly reconstructed reducing registered registration registrations replace required residual resolution rigid scanner scanning segmentation segmentations selected similarly slice software spline splines staring stochastic strategies stroke studies supported surface transactions transform transforms translation translational transverse truth tubular validated validation various vessel visualized wall weightings