Biomedical Engineering Reference
In-Depth Information
This restricted the application of MR-CT registration to small subsets of
patients for whom the decision to register the image was made in advance of
the acquisitions, and for whom it was acceptable to alter the image acquisi-
tion for the purposes of registration. In practice, this was rarely the case out-
side a small number of centers.
In the last few years, the difficulties caused by field of view have been
reduced because of the widespread use, first, of spiral (or helical) CT scanning
and, more recently, multislice* CT scanning. This has increased the number
of slices routinely acquired. Furthermore, it is becoming more common to carry
out 3D volume acquisitions with MR. These volume images tend to have
larger fields of view than the multislice acquisitions and have approximately
isotropic resolution, although they also tend to have fewer good contrast
characteristics. As a consequence, the volume of overlap between routinely
acquired MR and CT images of the same subject tends to be larger now than
it was ten years ago. Despite changes in image acquisition, it is still necessary
for a clinically usable MR-CT registration algorithm to be able to accurately
register images with different fields of view and slice orientations.
10.2.1.2
Resolution
The in-plane resolution of CT images tends to be higher than that of MR
images, as CT images are routinely acquired with an image matrix of 512
512 pixels, whereas MR images most commonly have a 256
256 image
matrix. Also, techniques used to speed up MR acquisition frequently result in
lower resolution in one in-plane direction (the phase encode direction) than
this matrix size suggests. In the through-plane direction, however, the situa-
tion is often reversed, with MR having higher through-plane resolution than
CT. The desire to minimize radiation dose in CT often results in the use of
fewer, thicker slices than are used in MR. In traditional multislice (one slice
at a time) CT acquisitions, the slice spacing can be changed part of the way
through an acquisition to minimize dose. For example, slices through the
skull base might be acquired with 3 mm slices and 3 mm slice spacing, but
slices higher in the head where less detail is required might be acquired with
5 mm slices and 5 mm slice spacing.
These resolution factors have important implications for MR-CT registration.
First, algorithms used for registration need to work with these differences in
resolution. Second, care needs to be taken when combining the registered
images, for example with a color overlay display. It is not possible to trans-
form the MR image to the coordinates of the CT scan, nor the CT image to
the coordinates of the MR scan, without degrading the resolution of the
transformed image. Given the effort put into the original data acquisition,
it seems wasteful if the registration process requires that one or another
* The latest generation of CT scanners can acquire about two or four slices simultaneously, rather
than acquiring multiple slices one at a time; the number of slices that can be acquired simulta-
neously is likely to increase as the technology matures over the next few years.
