Biomedical Engineering Reference
In-Depth Information
12.3
3D Image-to-Physical Space Registration
Having defined a coordinate system within the patient, it is now necessary to
align the preoperative images to this space. The process essentially consists
of identifying features in the preoperative images which can also be found on
the patient using the localization device.
While there are some similarities between image-to-image registration and
image-to-physical registration, there are specific problems associated with
image acquisition and identifying the physical features on the patient in the
operating room.
12.3.1
Preoperative Image Preparation and Planning
Care has to be taken in the acquisition of images for image guidance. The
images must be acquired with sufficient spatial resolution for the guid-
ance task. In practice, for CT and MRI this means that the slice thickness
must be sufficiently small. It is not unusual to need a slice thickness of 3 mm
or less in CT or MRI. The slice thicknesses of 5 mm or more in diagnos-
tic imaging will make it hard, if not impossible, to identify features
with the submillimetric accuracy necessary for certain image guidance
applications.
Geometric image calibration is required as part of the quality assurance of
the imaging device, and steps may need to be taken to reduce geometric dis-
tortions in MRI. CT gantry tilt should be known accurately and compensated
for if nonzero. Image guidance is one of the most demanding applications for
the geometric integrity of medical images. A 2% error in scaling over 200 mm
will result in a relative displacement of 4 mm, well outside the accuracy of
modern localizing devices. While CT and MRI manufacturers still quote
accuracies worse than this figure in their formal specifications, methods exist
to calibrate scanners with a high degree of accuracy.
20
While there are advantages in terms of reduced cost, radiation dose, and
time in using previously acquired images, in practice lack of access to the digital
data and the problems in image quality outlined above often mean that
repeat scans are required specifically for image guidance. This adds signifi-
cantly to the cost of image-guided procedures. Issues of image preparation
and artifact reduction are covered in more detail in Chapters 4 and 5.
Most image guidance entails a planning step, which may be as simple as
identifying a single target or may involve delineation and labeling of complete
structures such as a tumor in neurosurgery or individual bones in computer-
assisted orthopedic surgery. This process of segmentation is one of the weakest
links in image-guided-surgery. Software tools are improving, but fast and
accurate image segmentation software that requires minimal intervention is
still not available for most applications.
