Re: VBM and the hippocampus
For an example of VBM of medial temporal cortex see also: Shah, P.J., Ebmeier, K.P.,
Glabus M.F., Goodwin, G.M. (1998). Cortical grey matter
reductions associated with treatment resistant chronic unipolar
depression: a controlled MRI study. British Journal of
Psychiatry 172, 527-532.
We actually found the left medial temporal cortex reductions in
grey matter density to be correlated with verbal memory scores,
KPE
Dear Teri
Subtle grey matter changes can be detected in the hippocampi of
normal
subjects with VBM (which corroborate independent ROI measurements)
see
Maguire et al, PNAS, 2000. See also Salmond et al, HBM 2000 regarding
the
detection of bilateral hippocampal changes with VBM.
Mesial temporal sclerosis can be picked up on VBM using SPM 99
(unpublished
data) and hippocampal/amygdala changes can also be seen in elderly
diseased
patients (unpublished data)
VBM pre-processing has advanced since the Woermann et al paper,
particularly the spatial normalisation and segmentation steps, and
further
optimisations include modulation and automated brain extraction
(Ashburner
& Friston, VBM-the methods, NeuroImage, 2000 and Good CD et al,
NeuroImage
in press).
Although VBM is a fully automated procdure, attention to detail is
important. For example, I would suggest using customised
(disease/age/sex
matched) templates for spatial normalisation (which can also be
used as
the priors for segmentation).These methods are not perfect, and
inspection
of the segmented images for some elderly normal and diseased
patients
reveals areas of relatively poor tissue classification). Another point
to
be considered is the variance between subjects and between brain
regions in
the same subjects. In areas of increased variance VBM may be less
sensitive
to change, although this is the point of SPM which employs a
regionally
specific estimate of variance. The spatial normalisation in VBM does
not
aim to match each and every gyrus or deep brain structure, rather it
aims
to match overall brain shape. If the normalisation was perfect, there
would
be no differences between the normalised scans and all the
differences
would be in the deformation maps. In regions of high variance a voxel
of
interest may not represent exactly the same small structure for
each
subject in the group. High dimensional warps and TBM or corical
surface
mapping (Thompson et al, cerebral Cortex, 2001)are probably the best
way to
solve some of these problems, athough they may be time intensive.
I hope this helps
Tina Good
There have been At 20:54 25/02/01 -0500, you wrote:
>Can someone please explain to me why gray matter changes in
the amygdala
>(and hippocampus) cannot be identified using VBM and SPM99?
I have read the
>Woermann et al paper that describes visibly evident
hippocampal damage in
>epileptic patients that was not detected using VBM. The
authors reason that
>small abnormalities are excluded due to normalization,
segmentation, and
>smoothing. Although that makes sense to me - it doesn't make
sense that the
>hippocampal damage that could be seen with the naked eye could not
be
>detected with this method. They
also state that automatic segmentation and
>voxel-by-voxel comparison of
structural MRI is not suitable for
>investigating the amygdala and
hippocampus. Is it just these two regions?
>Why just these two regions? I would greatly
appreciate any references or
>help in resolving this issue. Thanks all, teri-)
>
>
Dr Catriona Good
Clinical lecturer / Neuroradiologist
Wellcome Dept of Cognitive Neurology, ION
12 Queen Square
London WC1N 3BG
email: [log in to unmask]
phone 0207 8337485