Image registration by maximization of combined mutual information and gradient information

Abstract

Mutual information has developed into an accurate measure for rigid and affine monomodality and multimodality image registration. The robustness of the measure is questionable, however. A possible reason for this is the absence of spatial information in the measure. The present paper proposes to include spatial information by combining mutual information with a term based on the image gradient of the images to be registered. The gradient term not only seeks to align locations of high gradient magnitude, but also aims for a similar orientation of the gradients at these locations. Results of combining both standard mutual information as well as a normalized measure are presented for rigid registration of three-dimensional clinical images [magnetic resonance (MR), computed tomography (CT), and positron emission tomography (PET)]. The results indicate that the combined measures yield a better registration function does mutual information or normalized mutual information per se. The registration functions are less sensitive to low sampling resolution, do not contain incorrect global maxima that are sometimes found in the mutual information function, and interpolation-induced local minima can be reduced. These characteristics yield the promise of more robust registration measures. The accuracy of the combined measures is similar to that of mutual information-based methods.

Maximum distance-gradient for robust image registration

“…Adaptation of mutual information by incorporating spatial information proposed by Pluim et al (2000) (hereafter referred as Pluim), which is the multiplication of the mutual information and a gradient term based on both the magnitude and orientation of image gradient. Second-order mutual information proposed by Rueckert et al (2000) (hereafter referred as Rueckert), which considers the co-occurrence of intensity pairs of adjacent points in each of the images to be registered.…”

“…Based on the results using clinical three-dimensional CT and T1-weighted MRI image volumes from the retrospective image registration evaluation (RIRE) project (West et al, 1997), it is experimentally shown that, at different image resolutions, the proposed method can give longer capture ranges than the conventional mutual information based method and two of its adaptations (respectively proposed by Pluim et al (2000) and Rueckert et al (2000)) with incorporating spatial information. This can obviously make the multi-resolution image registration more robust.…”

“…However, it has been suggested that the conventional mutual information based registration can result in misalignment in some cases (Penney et al, 1998;Pluim et al, 2000) and thus room for improvement exists. The conventional mutual information measure takes only image intensities into account, since by definition, it depends on the relative occurrence of intensities in individual images and co-occurrence of intensity pairs in both images.…”

“…By multiplying an external local gradient term, Pluim et al (2000) incorporated both orientation and magnitude of gradient into the conventional mutual information measure to ensure the alignment of locations of tissue transitions in images. The plots of the similarity measure as a function of misalignments around ground truth indicate that the registration function of the combined measure is smoother than that of the standard mutual information measure.…”

Deformable registration using edge‐preserving scale space for adaptive image‐guided radiation therapy

“…( 14 – 15 ) However, in addition to the high computational complexity of MI for 3D image, optimization process can always be trapped into local extremes when searching for the global optimum and, hence, result in misalignment. Although multiscale registration methods can improve the registration efficiency including speed, accuracy, and robustness, ( 5 , 16 ) Pluim et al ( 17 ) believed that one reason for misregistration in MI‐based registration is the absence of spatial information in MI measurement.…”