ETAC output interpretation

AFNI Message Board
1

Hello,

I ran a paired t-test using 3dttest++ with the -ETAC flag. The results in the {PREFIX}.{NAME}.ETACmask.global.{SIDE}.{FPR}.nii.gz file look great. This may be a fundamental misunderstanding on my part of ETAC, but is there any way to isolate anatomical regions into clusters from this binarized map? For example, with cluststim parameters I can threshold and then run 3dclusterize to obtain an ordered cluster map.

Is {PREFIX}.{NAME}.ETACmask.global.{SIDE}.{FPR}.nii.gz the “final” map and would I report the entire map as “one cluster” in a manuscript?

Thank you,
Cate

2

Hi Cate,

That binarized map is essentially the thresholded, clustered map one might get from 3dClusterize, just without the clusters ordered by size. If you want to get the cluster table and volume, go ahead and run 3dClusterize. For example, assuming that dataset is called ETAC.global.2sid.5perc.nii.gz, consider:

3dClusterize -ithr 0 -idat 0 -inset ETAC.global.2sid.5perc.nii.gz \
                     -NN 2 -1sided RIGHT_TAIL 0.5 -clust_nvox 2 \
                     -pref_map ETAC.clust.order.nii.gz

That will threshold that binary dataset at 0.5, use NN2 neighbors, and require at least 2 voxels for a cluster (since they are already clustered, 1 voxel is probably enough, though some programs might whine about that).

Does that seem reasonable?

To verify cluster counts, compare the output table against:

3dhistog ETAC.global.2sid.5perc.nii.gz
3dhistog ETAC.clust.order.nii.gz
  • rick
3

Hi Rick,

Thank you! I was able to obtain an ordered cluster map using 3dClusterize and have 75 clusters. Does that seem reasonable for ETAC output?

Here is the comparison using histog

3dhistog ETAC.global.2sid.5perc.nii.gz
#Magnitude Freq Cum_Freq
0 489222 489222
1 2078 491300

3dhistog ETAC.clust.order.nii.gz
#Magnitude Freq Cum_Freq
0 489222 489222
1 437 489659
2 218 489877
3 118 489995
…
74 5 491295
75 5 491300

4

Hello,

I am bumping this to see if this is typical output from ETAC?

Thank you,
Cate

5

Hi Cate,

While I do not have too much experience with ETAC numbers, those seem reasonable, with the only concern coming around the small clusters. You can check the single subject results in those areas to be sure the results are pretty consistent. But since this is based on permutation testing and not a minimum global cluster size, it is certainly feasible to get smaller clusters.

If you evaluate those small regions across subjects, do the results seem consistent?

  • rick