Neurosurgeons are confronted with a number of issues complicating their work in open-cranium interventions [19]:
arose from the need to address these issues.
Its role is to provide a surgeon with the means to visualize the image
of a hand-held probe in relation to the diagnostic image information
interactively during surgery. Although CT and MR images are
generally used, other modalities may also be integrated
[47,48,27] (e.g., PET, DSA, MRA, EEG, Video).
Additionally, stereoscopic images can provide depth perception, which
is useful for interpreting both DSA images and general 3-D volume
reconstructions [11,12].
Transformations involved in conventional stereotaxy map between two-dimensional and three-dimensional spaces. It is, however, possible to think about a set of sectional images as defining a volume, so the notion of a slice is then no longer essential. This makes it possible to find a three-dimensional transformation between the patient and image spaces. This ``volumetric'' approach has the advantage that only one transformation is required to link the two spaces, and there is no need to compute a different transformation for each slice. In the same way, the reference role of the reference stereotactic marker is no longer essential; it can be replaced by any structure (natural or artificial) visible on the patient and in the images. On the other hand, this approach makes assumptions about the relative position of the slices, which may not be rigorously respected in reality. For instance, it is assumed that the distance between slices is constant, and that there is no rotation between two contiguous slices, that the scaling from on slice to another is the same, etc. Lemieux et al. [38] have demonstrated that, in general, the three-dimensional transformation leads to a less accurate mapping of the points from one space to the other, when compared to a two-dimensional transformation computed on a slice-by-slice basis. This can be understood by the fact that a set of 2-D transformations can be seen as a piecewise linear transformation; it is then likely to provide a more accurate match in any given slice than a global transformation.