Jaap,
> Our initial thought was to use spm_est_smoothness.m & spm_resels.m Using
> a temporal lobe mask, we should be able to obtain a FWHM estimate for
> the temporal lobe area of each ADC-map, and after that a resel count
> for the temporal lobe. Then, we would apply a Bonferroni correction
> with the number of resels in stead of the number of voxels. However,
> if we try to obtain FWHM and RPV-file, using spm_est_smoothness on a
> normalized ADC map with a temporal lobe mask, we do not get reasonable
> FWHM estimates (the order of magnitude is 10^-30mm, in stead of an
> expected 5 mm). Therefore this doen't seem to work.
I'm not sure exactly why spm_est_smoothness is doing so badly. A
couple ideas
1. Make sure you have the leatest version of spm_est_smoothness,
2.7. Previous versions used sinc interpolation would would
certainly have problems with small VOIs. The current
version uses linear interpolation.
2. Analyze the whole brain, but then just confine your
inference to the temporal lobe VOI using the S.V.C. button
(where you can specify a mask image).
This approach, though, has the disadvantage of assuming
homogneous smoothness through out the brain. I've never seen
ADC residual images, but the FA residual images I saw were
very spatially structured and very non-stationary (had very
variable smoothness).
Another option, though, is to forego parametric Random Field methods
and use the data itself to find corrected thresholds. Where RFT
methods make assumptions about Gaussianity and smoothness to find
the distribution of the maximum statistic under the null, permutation
methods let you use the data itself to find an emperical distribution
of the maximum. (Permutation methods are implimented in SnPM,
http://www.sph.umich.edu/ni-stat/SnPM )
However, the one-patient-vs-group situation is one where it is difficult
to get enough permutations. If you have N normal controls and 1 patient,
there are only N possible unique ways to permute the data. Unless N is
large this won't be satisfactory, as permutation P-values are multiples
of 1 over the number of permutations.
If you want to compare a patient *population* to a normal population,
though, the permutation methods will definately work well for even
modest group sizes.
Hope this helps.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|