Hi,
On 9 Apr 2009, at 23:50, Ruchika Wadhwa wrote:
> Hello:
>
> I have been analyzing the 12 direction DTI data with patients and
> controls using TBSS. I
> used tfce for my final statistical analyses and get predicted
> results in the controls>patients
> contrast. However, when I use the cluster command to parse out the
> possible clusters in my
> data, from which I could pull out the mean FA values, I get this one
> big cluster that pretty
> much encompasses the entire contrast with some other clusters. The
> way I understand is
> that cluster works by combining all the areas with a similar p-value?
No - cluster produces clusters of _connected_ voxels above the
threshold chosen. So presumably for your chosen threshold (maybe 0.95
in the corrp image) things are all connected somehow across the
skeleton. If you want to generate more segregated clusters, you
could just choose a higher threshold e.g. 0.99?
But yes - local maxima would be another valid way of describing the
results. The --peakdist cluster option lets you change minimum
distance between peaks.
Cheers.
> Given that this cluster
> includes the genu of the CC, splenium and the posterior thalamic
> radiation amongst many
> others, it kind of defeats the purpose of pulling out mean FA from
> more defined clusters. Is
> there any other way that I can accomplish this?
>
> One way I was thinking of doing this would be to take the local
> maxima in the clusters and
> creating an ROI around that. But I was wondering what approach do
> people usually follow
> for making an ROI for DTI data? Is it just mostly creating a 10 mm
> box around the peak
> voxel? Any help would be much appreciated.
>
> Sincerely,
> -Ruchika
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|