Biomedical Engineering Reference
In-Depth Information
3.3.5.4
Applications
The flexibility introduced with manual selection of CPs makes the current soft-
ware suitable for non-rigid registration in many applications in addition to the
clinical procedures described in the Introduction. We have successfully applied
it to human MR-MR prostate images as shown here, rat CT-CT images, and CT-
PET lung images [50]. We believe that the registration method can be applied
to many organs other than the pelvis and prostate, multimodality images, and
intersubject images. In addition, we think it applicable to a variety of animal
experiments in which we are involved, including iMRI-guided thermal ablation
in pig and rabbit, prostate imaging studies in dog, and controlled drug release
studies in rat.
We conclude that our mutual information non-rigid registration is fast and
can be applied to a variety of applications. For prostate and pelvic imaging,
it works better than rigid body registration whenever the subject position or
condition is greatly changed between acquisitions. It will probably be a useful
tool for many applications in prostate diagnosis, staging, and therapy.
3.3.6
Summary
Many applications in prostate cancer management such as tumor localization,
possibly tumor staging, tumor targeting during therapy, assessment of adequate
treatment, and treatment follow up, require image registration of MRI volumes
and/or volumes from other imaging modalities. With regard to interventional MRI
guided radiofrequency thermal ablation for the minimally invasive treatment of
prostate cancer, registration applications include the comparison of registered
MR images acquired before and immediately after RF ablation to determine
whether a tumor is adequately treated. When images are acquired in different
patient positions and/or different conditions, the pelvis, prostate, bladder, and
rectum can deform and displace. Non-rigid registration is desired to correct for
such deformations.
We created a two-step, three-dimensional registration algorithm using mu-
tual information and thin plate spline warping for the prostate MR images.
First, automatic rigid body registration was used to capture the global trans-
formation. Features included a multiresolution approach, two similarity mea-
sures, and automatic restarting to avoid local minimums. Second, local non-rigid
