I was given a script created by an ex-colleague for performing a seed correlation analysis of resting-state fMRI data. However, after comparing it to the Simple Correlation tutorial on the AFNI website (
here), I'm not sure if I understand it correctly and would therefore like to double-check with you.
My input: individual subject time series (preprocessed: registration, normalization, nuisance parameter regression, smoothing, filtering)
The seed analysis is as follows.
For each subject:
- extract the mean time series from the seed region, resulting in a text file subjX_region.txt;
- calculate correlation map:
3dTcorr1D -prefix subjX_corrmap_r.nii -mask brain/mask.nii.gz subjectX/time/series.nii.gz subjX_region.txt
- perform a Fisher r-to-z transform:
3dcalc -a subjX_corrmap_r.nii -expr 'log((1+a)/(1-a))/2' -prefix subjX_corrmap_z.nii
Now I can perform the group analysis using the z-transformed correlation maps:
3dttest++ -prefix group_roi_analysis.nii -setA subj*_corrmap_z.nii
This results in a file with two sub-bricks. My colleague would then do:
3dcalc -a group_roi_analysis.nii[1] -expr a -prefix result.nii
and plot this onto a brain template.
However, reading the 3dttest++
documentation, it seems this sub-brick contains the t-statistic for each voxel; the corresponding mean z-score should be in the 0th sub-brick.
Do I understand it correctly that the sub-brick I should be plotting onto the brain template is actually the 0th sub-brick, possibly only for voxels whose t-statistic (read from the sub-brick [1]) is above a given significance threshold?
Thanks in advance!