> What MR sequence (T1, T2 , PD or EPI) to perform on a given subject to get
> an optimal segmentation into GM, WM and CSF maps using SPM99 templates
(T1,
> T2 , PD or EPI). Should we use a heavily weighted T1 sequence since SPM99
> favors T1 (the default).
> Is there an advantage of one sequence versus the other one or are they
> equivalent in essence given the segmentation algorithm implemented in
SPM99?
The quality of the segmentation that can be achieved is dependent upon many
factors, but most importantly on the contrast between the different tissue
classes, relative to the noise. Other factors include the effect of partial
volume and image non-uniformity (although the effects of this are reduced
in the SPM99b segmentation). It is also important that the image can be
registered well with the prior probability images, so should have a similar
contrast to the templates, or a linear combination of the templates.
The best segmentations from the FIL have been from T1 weighted images, but
a great deal of tweaking has been done to optimise the sequence (which we
hope to eventually publish). However, it is possible that other contrasts
could produce equally good segmentations. Another possibility is to use
dual echo images. By using information from both T2 and PD images, it is
possible to achieve better segmentation than by just using PD or T2 images
alone.
All the best,
-John
|