Motion detection and motion compensation for digital subtraction angiography image enhancement

Abstract

For the enhancement of digital subtraction angiography (DSA) images, registration of mask and contrast image prior to subtraction is a pre-requisite. One of the main requirements of this task is that the region-of-interest used for the calculation of the registration parameters should contain the vascular structures of interest. This, however, is also one of the main problems in DSA because the contrasted vascular structures can be regarded as a distortion that makes the images to be compared dissimilar. In this paper we present a comparison between three frequently used similarity measures and histogram-based similarity measures. This reveals the advantages of the latter. The data-driven approach is especially suitable for registration of two images which are identical except for some structures visible in one but not in the other image. Based on an energy similarity measure, a motion vector field is obtained by template matching, which gives a set of homologous landmarks or control points in the mask and contrast image. A point-based registration is performed fitting the parameter of an appropriate transformation for patient motion correction. An affine and an elastic transformation are compared for an abdominal fluoroscopic scene.