Print

Print
Hi Cristina,

Please, see below:


On 11 October 2016 at 11:25, C. Roman <[log in to unmask]> wrote:
Hi Anderson,

Thanks for confirming the correct inputs for fsl_glm and fslmaths. I have searched through the discussion board to find how to make a cluster map and pull values from the FA, or in this effect size map, but wanted to confirm I am on the right path for doing this.

If I just want to see the Cluster Index (i.e., list of significant clusters), I use:

cluster -i tbss_tfce_corrp_tstat1 -t 0.95
(This should pull up the number of voxels for each cluster, z-max, z-max x, etc.)

Yes, this gives a table with all clusters above 0.95.
 


To make an actual cluster map with a cluster size of 10, I have the following:

cluster --in=mytbss_tfce_corrp_tstat1 --thresh=.95 --osize=cluster_size_tstat1_95
fslmaths cluster_size_tstat1_95 -thr 10 -uthr 10 -bin cluster_tstat_95_10

Yes, this will keep then only the clusters that have size exactly equal to 10, not larger, not smaller.
 

Then to pull values from my effect size image, I have:

fslmeants -i cohen.nii.gz -m cluster_tstat_95_10 -o avgcohen_tstat1_95_10.txt
(This should provide a txt file of average effect sizes per significant region. Alternatively, I could pull the average FA values and calculate cohen's d by "hand" outside of FSL?)

This will give the average Cohen's d within the cluster. Another option is to use fslstats. Yet another is to average the signal within the ROI, then run the GLM, then compute the effect size.
 

Lastly, if I would like to pull a mean value for each label in an atlas file, I would use:

fslmaths myths_tfce_corrp_tstat1 -thr 0.95 -bin significant_results_mask
fslmeants -i all_FA_skeletonised -m significant_results_mask --label=$FSLDIR/data/atlases/JHU/JHU-ICBM-labels-1mm.nii.gz

I haven't used the option --label but from the description and from the content of the label file, it seems it does what you need.

All the best,

Anderson

 

Thank you so much!
Cristina