Thanks for sending in the overview. This might be a good point for me to
send in a note relating to the stats derived from such data. If people are
going to put this kind of data through FMRI-style analysis (including
inference with methods such as GRF like we currently have in FEAT) then
care needs to be taken wrt the validity of the final stats.
One point is obviously that the data needs to be reasonably Gaussian (and
at least a little spatially smooth for GRF to hold) - but I don't have any
reason to believe that the cross-subject "noise" here would seriously
violate this.
Of more concern is the interpretation of the stats in areas which are not
solidly one tissue-type across all subjects. Obviously if the underlying
values (such as FA from DTI or Magnetisation Transfer Ratio from MTR
imaging) depend on tissue type, then varying tissue type at a given
standard-space voxel across subjects will be a huge confound in the
cross-subject stats. This needs to be considered when interpreting the
results. One possibly approach to this (though not one we have applied in
great depth yet) was written up for MTR imaging for ISMRM last year - see
http://www.fmrib.ox.ac.uk/analysis/research/sam
Steve.
On Thu, 31 Oct 2002, Marenco, Stefano (NIMH) wrote:
> fractional anisotropy describes how anisotropic the movement of water is. In
> white matter, because of the hindering of water diffusion by the myelin
> sheaths and by the organization of tissue, FA is higher than in gray matter.
> MS patients will typically have low FA as compared to controls and high mean
> diffusivity in the plaque, but also in other areas of normal appearing white
> matter. In fact these reductions of FA have been correlated with disability
> in MS patients. Stefano
|