Registration of Whole-Body MR Angiography Images
This study focussed on imaging patients with asymptomatic peripheral arterial disease (PAD), using a fast, high-resolution, 5-station MRA technique wherein the body is sub-divided into 4 or 5 "stations" which are sequentially scanned by automatically moving the anatomy covered by each station into the isocentre of the magnet. Initially, pre-contrast images are acquired at each station, followed by the rapid injection of a bolus of contrast agent and a re-acquisition of all images while chasing the bolus down the body. The respective pre- and post-contrast images are subtracted to accentuate visualisation of the arterial system.
Although the subtraction of Pre- from post-contrast images accentuates the vessel visibility, a common problem with this approach is the mis-registration artefacts which inevitably occur due to patient motion, but also due to mis-alignment of the table following its movement in and out of the magnet between image acquisitions. Such mis-registration artefacts can cause significant deterioration of the resultant image quality, particularly evident when trying to visualise small vessels.
To address this problem, we used a novel image registration algorithm developed by our collaborator, Dr Zhuang (currently in the Shanghai Advanced Research Institute), to investigate improvements in image quality with this bolus-chasing technique. Image registration techniques are well established for dealing with patient movement artefacts and work particularly well in the brain. Conventionally, such techniques work through a direct analysis of pixel signal intensities or by identifying features within the two images to be registered, and hence a common assumption is that both images have the same signal intensities in corresponding anatomical structures which are simply displaced somewhat relative to each other However, such registration algorithms do not work well when contrast agents are used, since such agents are designed to specifically to change the localised signal intensity, whether in tumours in the case of dynamic contrast enhanced (DCE) studies. or blood vessels in CE-MRA examinations, rendering inaccurate the assumption of no signal changes between common image features the images to be registered. To address this problem, algorithms using normalised mutual information (NMI) have bene developed, and hence the aim of the study was to evaluate the efficacy of several such registration algorithms for the MRA data acquired during this study.
Example images: (a) non-registered subtracted images, (b) rigid registration, (c) affine registration, and (d) warping registration using the new algorithm
Test phantom developed to assess the ability of the registration algorithm to deal with large patient movement
- Foley D, Browne JE, Zhuang X, Sheane B, O'Driscoll D, Cannon D, Sheehy N, Meaney JF, Fagan AJ, "The utility of deformable image registration for small artery visualisation in contrast-enhanced whole body MR angiography", Physics Medica: European Journal of Medical Physics 30(8), 898-908, 2014, doi: 10.1016/j.ejmp.2014.08.001, PMID: 25182374