Improvement of vessel segmentation by elastically compensated patient motion in digital subtraction angiography images

Abstract

Digital subtraction angiography is a standard diagnosis tool for the examination of vessels. For this method X-ray images (contrast images) are taken from the patient while a radio-opaque contrast agent is injected through a catheter. The first image of such a scene is usually taken before injection and is called mask image. In clinical routine the mask image is manually shifted to perform a rough motion compensation. Then the corrected mask is subtracted from the contrast image to erase all disturbing permanent structures like e.g. bones or organs. In the vessel diagnosis chain a vessel segmentation is often applied to the subtraction results. Real DSA-images only corrected via an interactive-shift routine still suffer from motion artifacts which may lead to false results from the segmentation step. Especially, for the abdomen where the patient motion is very complex it is illustrated how residual artifacts result in mis-segmentations. In the present paper we demonstrate that an affine transformation, and particularly, an elastic transformation yield an excellent patient motion compensation which is a sufficient basis for the segmentation algorithm. We describe a registration procedure based on the estimation of a motion vector field. Additionally, we outline a new vessel segmentation algorithm.