Block Matching: A General Framework to Improve Robustness of Rigid Registration of Medical Images

Abstract

Abstract. In order to improve the robustness of rigid registration algorithms in various medical imaging problems, we propose in this article a general framework built on block matching strategies. This framework combines two stages in a multi-scale hierarchy. The first stage consists in finding for each block (or subregion) of the first image, the most similar subregion in the other image, using a similarity criterion which depends on the nature of the images. The second stage consists in finding the global rigid transformation which best explains most of these local correspondances. This is done with a robust procedure which allows up to 50% of false matches. We show that this approach, besides its simplicity, provides a robust and efficient way to rigidly register images in various situations. This includes for instance the alignment of 2D histological sections for the 3D reconstructions of trimmed organs and tissues, the automatic computation of the mid-sagittal plane in multimodal 3D images of the brain, and the multimodal registration of 3D CT and MR images of the brain. A quantitative evaluation of the results is provided for this last example, as well as a comparison with the classical approaches involving the minimization of a global measure of similarity based on Mutual Information or the Correlation Ratio. This shows a significant improvement of the robustness, for a comparable final accuracy. Although slightly more expensive in terms of computational requirements, the proposed approach can easily be implemented on a parallel architecture, which opens potentialities for real time applications using a large number of processors.

A comparison of voxel and surface based cortical thickness estimation methods

“…In these experiments, we register the AAL atlas (Tzourio-Mazoyer et al, 2002) to each subject using block matching (Ourselin et al, 2000) followed by a spline based non-linear deformation Rueckert et al, 1999) both implemented in NiftyReg 4 , or alternatively we use the FreeSurfer parcellation to define the regions. For each subject the GM mask is assigned region labels based on the closest atlas or parcellation label.…”

“…These were chosen as they are available in both the FreeSurfer and AAL atlases, and of interest in these neurodegenerative diseases. The AAL atlas was registered to the T1w volume using block matching (Ourselin et al, 2000) followed by a spline based non-linear registration Rueckert et al, 1999). For each of the 49 subjects in cohort 1, and each atlas, the mean cortical thickness of over each atlas region was calculated as described in section 2.4.…”

A three-dimensional, histological and deformable atlas of the human basal ganglia. I. Atlas construction based on immunohistochemical and MRI data

“…Principal features of Yav++ include comparison and fusion of 3D images in multiplanar viewers as well as a 3D camera allowing visualization of serial contours, 3D surfaces and 3D images in the same image (see Appendix B for details). The BALADIN software, an automatic image registration algorithm, was also developed by the Epidaure group at INRIA (Ourselin et al, 2000). It allows registration of 2D or 3D images through an intensity-based robust block-matching approach (technical details in Appendix C).…”

Atlas-based delineation of lymph node levels in head and neck computed tomography images

“…• a global a ne transformation is computed between P andM S , based on a robust Block-Matching registration algorithm [28], • then, the remaining local deformations due to inter-patient variability are recovered using a non linear registration method.…”

“…This method, also presented in [29], is an extension to dense transformation of Block-Matching based rigid registration [28]. At each iteration i, pairings are computed between the images using Block-Matching.…”