Biomedical Engineering Reference
In-Depth Information
To achieve results in a time frame that is useful during surgery, how-
ever, a simpler model is likely to be required. For spinal surgery, a method
which provides a smooth deformation between the rigid vertebrae has
been proposed.
49,50
To incorporate fluid regions we have also proposed a sim-
plified tissue model with rigid, fluid, and deformable components.
51
In this active area of research, it is important to consider the clinical applica-
tion when developing an algorithm. The tissue model does not necessarily
have to be physically accurate. Intraoperative imaging can provide much
input data about the current position and shape of the patient's anatomy, and
this information may be sufficient to constrain a much simplified tissue
model. The model also needs to be only as accurate as the application requires.
An algorithm based on approximating B-splines
52
has been used to align
immediate pre- and post-intervention MR images for a range of neurosurgical
procedures. The resulting deformation fields have been used to determine how
much brain tissue has distorted or moved as a result of the intervention.
53
Ultrasound is being developed as less expensive alternative intraoperative
imaging modality to update preoperative images. A method has been pro-
posed that involves extraction of contours from the B-scans, which are con-
verted to deformation vectors that are linearly interpolated throughout the
volume.
54
This method provided registration accuracies of better than 2 mm
on a phantom designed to mimic the brain and ventricular structures for
image-guided neurosurgery. King et al.
55
have recently proposed a Bayesian
method to use intraoperative ultrasound to update the location of contours
of structures delineated in preoperative MR images.
12.6
Clinical Applications
Registration of images to the patient in the treatment room provides the
clinician with accurately aligned data to ensure that treatment is given in
accordance with the preoperative plan. Benefits to the patient include
smaller entry wounds (e.g., craniotomies), lower failure rate (e.g., pedicle
screw placement), and less morbidity, due to avoidance of critical struc-
tures. Current applications of preoperative image alignment to the patient
tend to assume the rigid-body transformation. The registration accuracy is
limited by the amount of soft tissue deformation. This limits surgical appli-
cations to those near bone, such as skull base surgery,
56
spinal surgery,
45,46
or orthopedics.
21
Alignment of images to the patient using stereotactic frames has long been
a part of neurosurgery. It is in the field of neurosurgery that the first frameless
surgical navigation device was developed.
11
The first system in regular
clinical use was the mechanical version of the ISG viewing wand. A range of
commercial systems is now available, most of which use optical tracking.
A number of clinical trials have established the efficacy of image guidance in
