Biomedical Engineering Reference
In-Depth Information
These methods of matching to bone assume the rigid-body transformation.
Ultrasound also provides potentially useful data about the position of soft
tissue structures deep below the operative surface that have deformed since
the preoperative scan (see Section 12.5).
12.4.4
Interventional MRI
With interventional MRI (iMRI) a fast-imaging modality with 3D capabilities
is available in the operating room. One may ask why registration is necessary
if we have an instant 3D image of the patient. There can be, however, much
information in preoperative scans unlikely to be available from iMRI, for
example functional data or the precise location of bone. Also, the quality of
preoperative images from a diagnostic scanner is likely to be much higher in
terms of contrast, signal-to-noise ratio, and geometric distortion.
Tracking of instruments relative to the most recent iMRI scan is also an issue
that may involve a registration process. The iMRI scans could be taken at reg-
ular intervals throughout the procedure, with real time tracking of a pointer
or tool to provide guidance in between the scans. This would most likely
include marking of fiducials or surface points as described in Section 12.3. If
the procedure is performed entirely within the scanner and the patient does
not move, it is also possible that such alignment could be achieved by scanner
calibration.
Volumetric intensity-based registration, especially with the use of mutual
information as a similarity measure, has become the most popular and robust
method for alignment of multiple 3D images.
43
It seems likely that such
methods will also prove useful in the operating theater for alignment of pre-
operative scans to iMRI.
47
iMRI should prove particularly useful in generat-
ing updated 3D images which compensate for soft tissue motion during
interventions (see Section 12.5).
12.5
Tissue Deformation Correction
Nonrigid registration is the subject of Chapters 13 to 15. Image to physical
registration for therapy guidance poses specific problems in this area. If the
surgery or treatment plan is created using a preoperative image, this should
ideally be updated according to any movement of the patient, whether rigid
or nonrigid.
One way of tackling this problem is to create a physical model that closely
matches the tissue properties, measure only the required physical parameters
to constrain this model, and update the registration accordingly. Use of the
finite element method to create approximations to tissue mechanics, for
example by using linear elastic elements, has been attempted.
48
Up to 70% of
the deformation of the brain could be recovered merely by using the direction
of gravity as input. This approach is described in greater detail in Chapter 15.
