Biomedical Engineering Reference
In-Depth Information
our study are consistent with earlier reports on respiration-induced prostate
movement of
≤
1 mm for most patients in supine position with “quiet” respira-
tion [32]. The difference between the treatment and diagnostic positions resulted
in the most consistent and largest displacement of the prostate. When images
were acquired in the diagnostic position one week apart, there was significant
displacement of the prostate due to a change in rectal filling. This is consis-
tent with previously reported results [8, 14], which found rectal filling to be a
significant factor in prostate displacement.
There are ways to limit the small displacements of the prostate. One obvious
remedy is to acquire images in the same position. That is, if we want to register
an image with one obtained in the treatment position, we should obtain it in
the treatment position. Although it is unknown how accurately one must repeat
the treatment position, a device to support and constrain the legs is probably
required. In addition, there is a dependence of registration error on bladder
and rectum content. One solution is use clinical preparations often employed to
void the bladder and rectum prior to prostate imaging or therapy. We anticipate
that this might even lessen prostate displacements between the diagnostic and
treatment positions.
We must consider our results with regard to potential applications such
as those described in section 3.1. First, registered images acquired before and
immediately after treatment can be used to determine whether a tumor is ad-
equately treated. Second, serial examinations can be registered to determine
tumor progression or regression. Third, registration of functional images from
other modalities such as nuclear medicine or from MR spectroscopy can give
molecular markers for prostate cancer [33, 34]. Fourth, we want to register high
quality MR images with a few live-time interventional MR images to aid treatment
decisions [4, 5]. Our results indicate that registering images from the treatment
and diagnostic positions can lead to errors, and potential steps are described
above to limit this error. With images acquired in the same position, our results
place a lower limit on registration error of about 1 voxel.
3.2.5.2
Assessment of Registration
We are involved in a long-term effort to use registration for detection, assess-
ment, and therapy of prostate cancer. Hence, we have developed and used sev-
eral methods to assess pelvic and prostate registration.
