Hi Anderson
I think I found out the reason of the discrepancy between what you said the behavior of randomise with a mask should be and the outputs I get from implementing the TFCE with FEAT GUI by selecting a prethreshold mask.
First, regarding the masking issue, the "pre-threshold mask" selected from GUI was ever only used to mask the zstat, and that is why the rendered_thresh_zstat did show the activation limiting to the cluster.
However, the randomise command call from the log in the FEAT report showed the following:
/usr/local/fsl/bin/randomise -i filtered_func_data -o stats/ -m mask -d design.mat -t design.con -e design.grp -n 5000 -T --film
That is, it only used the "mask", which was the intersection of all the subjects' masks, rather than the "pre-threshold mask" specifically selected. So this behavior is different to when one use the command line randomise directly and specify -m Accumbens.nii.gz for example. So that is why the ouput from FEAT GUI with TFCE led me to think that the TFCE was calculated the same with or without masking. (As the matter of fact, they were.)
Secondly, the cluster report was generated by FEAT from the following command:
/usr/local/fsl/bin/cluster -i thresh_zstat2 -t 0.95 --othresh=thresh_zstat2 -o cluster_mask_zstat2 --connectivity=26 --mm --olmax=lmax_zstat2_std.txt --scalarname=Z -p 0.05 -d 1 --volume=915 -c stats/cope2 --empiricalNull=thresh_pstat2 > cluster_zstat2_std.txt
After several trail and error, I found out that what actually limit the spatial content to the accumbens ROI is by masking the input to the --empiricalNull flag with the accumbens mask. The "thresh_zstat" was already masked by the accumbens mask as showed by the log, however the thresh_pstat was only ever masked by "mask", not accumbens_mask, and this cause the cluster command to find out cluster within the masked region of the thresh_pstat, which is the whole intersection mask.
When I masked thresh_pstat with the accumbens mask, the following is what I get from the cluster report:
Cluster Index Voxels P -log10(P) Z-MAX Z-MAX X (mm) Z-MAX Y (mm) Z-MAX Z (mm) Z-COG X (mm) Z-COG Y (mm) Z-COG Z (mm) COPE-MAX COPE-MAX X (mm) COPE-MAX Y (mm) COPE-MAX Z (mm) COPE-MEAN
2 240 0.0026 2.59 3.54 -14 10 -10 -11.6 11.2 -5.39 48.4 -8 8 0 22.4
1 239 0.0026 2.59 3.35 12 16 -4 10.8 14 -4.03 47.5 8 8 0 21.6
Which made much more sense.
Therefore, I think it is better to use command line randomise directly for now?
Thank you very much!
Chen-Chia
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|