Print

---

Hi Archana

It seems the differences is highly distributed across the white matter tracts. This is fine and a very interesting results indeed. Just make sure you included all needed covariates in the design model. If this is the result even after FWE correction, then great. I’m usually very suspicious of my data when I get such big differences in white matter, but it happens.

I see what you want to do, the problem is that you got a big cluster and it will not show any cut-off point.
The way Anderson says is actually a good way to do it, to run the t-tests for each ROI.

What you could also do, and I’m not sure how correct is this, is to use each tract as a mask (you can get the masks the white matter atlas), binarize them, and use them to extract the t-values and p-values using fslmaths. The problem is, each tract will have only one big cluster, but then you would already have the anatomical location.

Perhaps someone else has another opinion.

Hope this helps

Kind regards

Eduardo


> On Mar 16, 2015, at 3:19 AM, Archana Malagi <[log in to unmask]> wrote:
> 
> Hi Eduardo,
> 
> Yes, i ran tbss_fill. I want to study white matter differences in two
> language groups using dti metrics. For group comparison, i used two sample
> t-test with 10,000 permutations using randomise.
> 
> How can i use cluster command with white matter mask?
> 
> Thanks,
> Archana
> 
> 
> 
> # Hi Archana
> #
> # The map looks odd to me. Too many areas significant. Did you inflate the
> # results using fill? Can you elaborate on the study and the design?
> #
> # Another thing you can do is to extract the p and t values from the
> # clusters
> # using the white matter masks for each white matter tract.
> #
> # Eduardo
> #
> # El Sunday, March 15, 2015, Archana Malagi <[log in to unmask]> escribió:
> #
> #> ---------------------------- Original Message
> #> ----------------------------
> #> Subject: Re: [FSL] How to extract cluster size and p values
> #> From:    [log in to unmask] <javascript:;>
> #> Date:    Mon, March 16, 2015 10:31 am
> #> To:      [log in to unmask] <javascript:;>
> #> --------------------------------------------------------------------------
> #>
> #> Hi Rosalia and Eduardo,
> #>
> #> As discussed, i applied threshold 0.96 , 0.97 while using cluster
> #> command
> #> and also attached snapshot of tbss results :
> #>
> #> threshold 0.96
> #>
> #> Cluster Index   Voxels  MAX     MAX X (vox)     MAX Y (vox)     MAX Z
> #> (vox)     COG X
> #> (vox)   COG Y (vox)     COG Z (vox)
> #> 1       32483   0.999   70      150     105     84.3    126     94.2
> #>
> #> and threshold 0.97
> #>
> #> Cluster Index   Voxels  MAX     MAX X (vox)     MAX Y (vox)     MAX Z
> #> (vox)     COG X
> #> (vox)   COG Y (vox)     COG Z (vox)
> #> 1       31264   0.999   70      150     105     84.1    126     94.2
> #>
> #>
> #> Results of atlasquery with threshold: 0.96 and 0.97 gives same results.
> #>
> #>
> #> As a result, it gives only one cluster for both the thresholds and most
> #> of
> #> cluster the lies is in right hemisphere.
> #>
> #> Thank you,
> #>
> #> Archana
> #>
> #
> #
> # --
> # Eduardo A. Garza Villarreal, M.D., Ph.D.
> #
> #
> #    - Investigador, *Subdirección de Investigaciones Clínicas, Instituto
> #    Nacional de Psiquiatría.*
> #    - Catedrático, *Consejo Nacional de Ciencia y Tecnología (CONACYT).*
> #    - Investigador Asociado, *Center of Functionally Integrative
> #    Neuroscience, Universidad de Aarhus, Dinamarca.*
> #    - *Profesor-Investigador, Servicio de Neurología, UANL.*
> #    -
> # *Coordinador, Unidad de Neurociencias, Centro de Investigación y
> # Desarrollo
> #    en Ciencias de la Salud, UANL. *
> #