Sph

The assumption of a linear relationship between voxel values is also, in general, untrue for images of different modalities, as is

1This strict requirement is seldom true in intramodality registration, either, as noise in medical images such as modulus MRI scans is frequently not Gaussian, and also because there is likely to have been a change in the object being imaged between acquisitions.

Here, (a) and aB(a) are the mean and standard deviations of the values of voxels in image B that co-occur with value a in image A, nA(a) is the number of voxels with intensity a in image A, and N is the total number of voxels, and similarly for ^A(b), cA(b), and nA(b). Although the idealized assumption made by the algorithm can be criticized as too simplistic, this approach proved to be extremely successful and led to renewed effort from the medical image analysis community to find alternative intensity-based algorithms that would work for

FIGURE 1 Example unregistered MR (left) and CT (right) images. The top panels are the original axial images, and the lower panel the reformatted coronal images. Note the striking difference in intensity of equivalent anatomical tissues between these images. Also, for these scans targeted at the temporal bone, the CT images have a much smaller axial field of view than the MR images. The difference in field of view of images being registered is one of the greatest challenges of intermodality image registration.

FIGURE 1 Example unregistered MR (left) and CT (right) images. The top panels are the original axial images, and the lower panel the reformatted coronal images. Note the striking difference in intensity of equivalent anatomical tissues between these images. Also, for these scans targeted at the temporal bone, the CT images have a much smaller axial field of view than the MR images. The difference in field of view of images being registered is one of the greatest challenges of intermodality image registration.

other types of images (e.g., MRI and CT), or that would remove the need to presegment the MR image data. One tool used in devising new similarity measures was the two-dimensional or joint image histogram.

0 0

Post a comment