Dear experts,
I'm now doing some task-fMRI analysis about the group difference between internet disorders and health controls. Since no regions of ROI could survived FDR correction for multiple comparisons using a small volume correction, I've tried the REST's AlphaSim program for whole-brain and small volume correction. And I'm not sure whether my manipulation was right, and I’d like to ask for your advice.
1. when doing FWHM estimation in Rest Alphasim, I chose the mask corresponding to the Statistical Map (parts of parietal lobe were chopped out due to the volumes scanned with an orientation of 30o from the AC-PC line), and the result of FWHM was >15 with the correction at p<0.05 for multiple comparison (voxel level p<0.01, rmm = 5 mm), which revealed a cluster size of 278 contiguous voxels.It seems that the cluster size was quite big. I wonder if it is correct to use the BrainMask_05_61x73x61, since when I used this mask instead, the FWHM was just 12 and the cluster size was 160 voxel.
In addition, I have also tried no mask (no selected file in mask option while estimating FWHM). Surprisingly, the FWHM was only 10.Would you give me some advices about what kind of mask should I choose when doing FWHM-estimation (mask for statistical maps, 61x73x61mask, or no mask)?
2. And what about the SVC, is it right to choose the ROI-mask in doing the AlphaSim program? Should I also choose the ROI-mask when doing the FWHM-estimation?
3. Another issue bothered me was that when I run FWHM estimation with no mask, the volume displayed in Matlab command window was 271633 (61x73x61), whereas after the rest AlphaSim with the BrainMask_05_61x73x61,the volume displayed in the Alphasimtext was 70831. I wonder if the latter means the volume of the brain?
Thank you for your time and consideration.
Best,
Vincent
Hi Vincent,
Hi Vincent,
1. An explanation is that the outside brain noise is less smooth as compared with the signals inside brain. Thus, when you use no mask, the FWMH is smaller. I will recommend you to use the group mask generated in DPABI->QC, if you have used DPARSF 3 and up to process your data. Also, use the 4d residuals to estimate the smoothness will be better (DPABI->Statistical analysis will do that automatically). BrainMask_05_61x73x61 is acceptable, howvere, you may have some noisy results around the brain edge, if the coverage is smaller accross your subjects.
2. For SVC, if you have a hypothesis, you can define a mask, then perform FWHM-estimation and correction within that mask.
3. 271633 is the number of total voxels. 70831 is the number of voxels within the brain.
You can also try GRF correction. Please see more details in http://rfmri.org/Course V2.0 Part I.
Best,
Chao-Gan