Image Loading Problems and Limitations
In order to register 3D image volumes from different modalities, we cannot
just manipulate raw pixel or voxel values. For correct and automated registration
rview generally needs to know extra information which relates voxel values
to locations in 3D space within the patient:
1. Voxel Size/Spacing
Derived from field of view, slice thickness, zoom factor, slice separation
etc.
Note: Voxel dimensions as aqcuired by many scanners are only accurate
to within a few
percent: this is particularly relevant for:
-
All MR data
-
Spiral/Helical CT data (variations in bed speed)
This may have a significant effect on the accuracy of a global rigid
registration.There are two approaches to this:
a) Use rview to do a scale estimate
This only gives relative correction between the two scans not an absolute
estimate of voxel size which will be required if, for example, the images
are to be used in surgical guidance etc.
b) Measure the true voxel dimensions independently in the scanner by, for
example, using a phantom.
2. Slicing direction or orientation:
Transaxial, Sagittal, Coronal:
rview cannot in general automatically align slices acquired in different
planes (for example transaxial to coronal) without some form of starting
estimate either derived from slice orientation (read from the scanner data
file) or from manual input.
Reflection: An even harder problem is to automatically correct refections,
since medical image data can be highly symmetric (eg left right swap).
It may be possible to develop a system to test all possible refelctions,
but it is far safer and quicker to ensure this information is provided
by the scanner data.
3. Gantry Tilt:
This is an additional problem posed by CT where a gantry tilt can often
be used to target the CT slice orientation. rview currently assume the
gantry tilt is zero if it cannot read this from the file.Note: Automated
reading of grantry tilt from All CT files is not yet fully supported by
rview.
When not using raw image data files, but data files
from other image processing environments (e.g. analyze, gipl), it is VERY
important to ensure the above information has been converted and stored
correctly in the intermediate format.