We are looking at the inhomogeneity correction of t1 image in the
segmentation part of SPM99. Does anybody know why there is about 10%
difference in the corrected volume between using stripped (of non-brain
materials) and unstripped input? In both cases, we chose the "spatial
normalization" (input is not spatially normalized) and "a little bias" option.
The difference has some structures in it and does not appear to be due to
round off errors.
Thanks.
Jeih-San Liow
|