Hi-
Thanks for sending the mask. I show it overlaid on the one I had been testing---it is indeed a bit narrower, esp around the ventricles. I don't know that that is an inherent problem, but those narrow regions are kind of non-ideal for the 3D cluster calcs and smoothing. It might just be part of life when using a GM mask. When I ran with just the default 10,000 iterations, the cluster sizes were 45, 47 and 49 (see below for full output).
I am not sure why you are switching to a different voxel size in that other mask? Note that regridding does introduce blur/smoothing into data, and that may change things slighty in your final analysis.
It looks like the cluster size has a reasonable convergence around 47+/a couple. Note that in all these things, it is good to keep an eye on the main goal: presenting reasonable results. If results appear/disappear because of having 47 vs 48 voxel thresholding, that is a sign of deeper instability in analyses. Reducing the sensitivity to these kinds of thresholds is discussed in this draft, with transparent thresholding:
https://www.biorxiv.org/content/10.1101/2022.10.26.513929v2
and here with ETAC:
https://pubmed.ncbi.nlm.nih.gov/31115252/
--pt
# 3dClustSim -mask MNI152_T1_2mm_brain_GM_02182017.nii.gz -acf 0.5128 3.8451 11.3839 -nodec -prefix TMP_CSIM_TEST2a
# bi-sided thresholding
# Grid: 91x109x91 2.00x2.00x2.00 mm^3 (160742 voxels in mask)
#
# CLUSTER SIZE THRESHOLD(pthr,alpha) in Voxels
# -NN 2 | alpha = Prob(Cluster >= given size)
# pthr | .10000 .05000 .02000 .01000
# ------ | ------ ------ ------ ------
0.050000 694 828 1024 1203
0.020000 268 332 415 477
0.010000 155 190 242 277
0.005000 93 116 147 173
0.002000 54 66 84 100
0.001000 36 45 59 70
0.000500 25 32 42 52
0.000200 15 20 27 34
0.000100 11 14 19 24
# 3dClustSim -mask MNI152_T1_2mm_brain_GM_02182017.nii.gz -acf 0.5128 3.8451 11.3839 -nodec -prefix TMP_CSIM_TEST2b
# bi-sided thresholding
# Grid: 91x109x91 2.00x2.00x2.00 mm^3 (160742 voxels in mask)
#
# CLUSTER SIZE THRESHOLD(pthr,alpha) in Voxels
# -NN 2 | alpha = Prob(Cluster >= given size)
# pthr | .10000 .05000 .02000 .01000
# ------ | ------ ------ ------ ------
0.050000 728 881 1088 1262
0.020000 287 351 441 505
0.010000 164 202 251 287
0.005000 101 124 157 183
0.002000 58 72 91 108
0.001000 39 49 62 73
0.000500 27 34 44 52
0.000200 17 22 28 34
0.000100 12 16 21 26
# 3dClustSim -mask MNI152_T1_2mm_brain_GM_02182017.nii.gz -acf 0.5128 3.8451 11.3839 -nodec -prefix TMP_CSIM_TEST2c
# bi-sided thresholding
# Grid: 91x109x91 2.00x2.00x2.00 mm^3 (160742 voxels in mask)
#
# CLUSTER SIZE THRESHOLD(pthr,alpha) in Voxels
# -NN 2 | alpha = Prob(Cluster >= given size)
# pthr | .10000 .05000 .02000 .01000
# ------ | ------ ------ ------ ------
0.050000 698 851 1058 1223
0.020000 277 335 434 505
0.010000 156 194 247 292
0.005000 96 120 153 177
0.002000 55 68 88 105
0.001000 37 47 61 72
0.000500 26 33 43 51
0.000200 16 21 28 34
0.000100 11 15 20 25
Attachments:
open |
download -
img_mask_compare.jpg
(177 KB)