Cyril Pernet wrote:
> Hi Guys,
> Thank you for your comments...
>
>> I was told that it is not recommended to use data from different
>> scanners for VBM. Differences among groups may be due to the
>> different scanners, or you may get noisy results. Worth checking this
>> before starting...
>
>
> I agree with you Mauro! In principle, if gp1 = scanner A and group2 =
> scanner B, then you have confounding effects between groups and
> scanners, and the analysis is not possible (I think). However, as 1/6 of
> the data (gp1&2) comes from a scanner A, 1/6 from a scanner B and 2/3
> from a scanner C, i.e. the repartition of subjects from gp 1 & 2 is
> approximately the same for each scanner. I don't think that future
> differences between groups (I hope) would be due to differences between
> scanners. First, because I will look at differences related to
> differences in neuropsy data and second because the difference between
> scanners will be modelled in the design matrix (as suggested by John).
>
> Why do you think that my data would be noisier than with all data coming
> from the same scanner?
>
>> The segmentation procedure shouldn't have problem with that kind of
>> data, should it ?
>
>
> Regarding the segmentation, it should be OK. The 1st segmentation would
> give different results (it is the case indeed) but the normalization on
> MNI T1GM should resolve the pblm of different sizes.
OK, you use the modified protocol - that was not specified in your first
mail -.
I'm not quite confident with this protocol, first due to the
segmentation in different spaces, and with the normalization on the GM
with the prior. In the 2001 Good's publication, it was said that a paper
would be published on the validation on this protocol. Do somebody know
if it was done ?
> Best,
> Cyril
>
>
|