Towards a Better Comprehension of Similarity Measures Used in Medical Image Registration

Abstract

Abstract. While intensity-based similarity measures are increasingly used for medical image registration, they often rely on implicit assumptions regarding the physics of imaging. The motivation of this paper is to determine what are the assumptions corresponding to a number of popular similarity measures in order to better understand their use, and finally help choosing the one which is the most appropriate for a given class of problems. After formalizing registration based on general image acquisition models, we show that the search for an optimal measure can be cast into a maximum likelihood estimation problem. We then derive similarity measures corresponding to different modeling assumptions and retrieve some well-known measures (correlation coefficient, correlation ratio, mutual information). Finally, we present results of registration between 3D MR and 3D Ultrasound images to illustrate the importance of choosing an appropriate similarity measure.

Block-matching strategies for rigid registration of multimodal medical images

“…sum of squared differences, correlation coefficient) or multimodal (e.g. mutual information, correlation ratio) problems [9].…”

Image registration and atlas-based segmentation of cardiac outflow velocity profiles

“…C = C(T ), so to avoid influence of the transformation function on the similarity measure. The problem of non-existent values for the source image is solved as suggested by Roche et al in [40]. In short, these values are artificially generated during the transformation, using the pixels from the image border.…”

Using recurrent neuro-fuzzy techniques for the identification and simulation of dynamic systems

“…Therefore, a visual convincing deformation is di$cult to obtain and a "rst initialisation with a fuzzy system did not lead to a su$cient result [2]. Therefore, a very promising possibility developed as part of the project ROBOSCOPE can be used to measure the deformation of the brain tissue: the coregistration of real-time 3D ultrasound and pre-operative MR data sets [34]. In ROBOSCOPE this is used to match the cheep and low-quality 3D ultrasound data sets with the high-quality, very expensive and time-consuming MR data sets, which are created before the operation.…”