Model space can be defined as the digital representation of the two physical volumes that can be processed by a computer. The links between physical and model spaces are provided by some measurement processes. Three factors limit the extent to which the physical volumes are accurately represented in model space.
The first goal of any Image-Guided neurosurgical procedure is to
estimate the transformation between the volumes in physical
space by using the imperfect information available in the model
space. The estimated transformation
will in general be
different from
because of the errors (mentioned above)
introduced in the transformation from physical space to model space.
Another error between
and
is introduced by the fact
that the non-linear part of
is not usually taken into
consideration in the search for
. In other words,
is chosen to be the rigid body transformation that best matches
. The error caused by this effect increases as the non-linear
part of
(i.e., the deformation between the two volumes)
becomes more important. To take the non-linear part of the
transformation into account would require the use of non-linear
registration techniques
[4,66,24,25]. We note
here that the main factor contributing to the difference in shape of
the patient's head between the scanner space the surgical space are
swelling, displacement of tissues due to gravity (since the position
of the patient is not necessarily the same in scanner and OR) and
expansion or contraction of the brain after craniotomy due to pressure
and drug-related effects.
It may not seem obvious why one should be concerned about the deformation of the skin since the region of interest is within the brain. However, registration must be performed before craniotomy and the points to determine the transformation matrix are generally situated on the skin.