Voxel-Based Similarity Measures for Medical Image Registration in Radiological Diagnosis and Image Guided Surgery

Abstract

Registration of images is a key technique for numerous medical applications from diagnosis to image guided therapy. In the present paper we focus on gray-value based registration methods. Of special interest is the so-called similarity measure which must be optimized. It is the intention of the paper to emphasize the fact that a successful registration requires a similarity measure that is carefully chosen with respect to the underlying medical application. Two single-modality examples are presented in the paper, i.e. the diagnostic tool of digital subtraction angiography and an intervention guidance under X-ray fluoroscopic control. We discuss new similarity measures, i.e. the class of one-dimensional hi stogram based measures and the pattern intensity, designed for these applications and compare them with frequently used measures like the cross-correlation function , cross-structure function and deterministic sign change criterion. It is demonstrated that the results obtained with the new measures are superior to the results obtained with the latter mentioned ones.