Hi,

I’m trying to get voxelwise correlations between two 4D nifti images. In my GLM, Y is a 4D image of 18 COPE images from task fMRI; X is a 4D image of 18 images derived from resting state fMRI. I would like to see if a subject’s resting state data is related to their stimulus response by getting a voxelwise correlation between the different image types.

I have successfully been able to get PALM to do a regression between the image types, and it produces the regression coefficient map (palm_vox_cope.nii.gz) and its associated t-statistic map. (palm_vox_tstat.nii.gz)  However, I would also like to be able to get a correlation coefficient map in order to derive an r-squared map (palm_vox_rsqstat.nii.gz) . In other non-voxelwise palm commands, I can simply add in the “–pearson” flag, but this doesn’t work for my voxelwise command. I get the following error:

Error using palm_takeargs (/usr/local/palm/palm_takeargs.m:1196)
The option "-evperdat" is incompatible with the options listed below:
"-ev4vg"
"-pearson"

Error in palm_core (/usr/local/palm/palm_core.m:33->palm_takeargs)
Error in palm (/usr/local/palm/palm.m:81->palm_core)

I then tried to get an f-statistic map using the “-f design.fts” and “–fonly” flags, from which I was going to derive the r-squared map using the following formula:

R-squared = (F * DOF1)/( F * DOF1 + DOF2), where DOF1 = rank(contrast matrix) and DOF2 = N – rank(design matrix).

However, when I add the “-f” flag alone, I get two tstat images instead of one tstat and one fstat image (I get palm_vox_tstat_c1.nii.gz and palm_vox_tstat_c2.nii.gz). If I use both “-f” and “-fonly”, I get no images back without an error.

I have also tried using the COPE output to generate the r-squared image using the following steps:

fslmaths X.nii.gz –Tstd Tstd_X.nii.gz
fslmaths Y.nii.gz –Tstd Tstd_Y.nii.gz
fslmaths palm_vox_cope.nii.gz –mul Tstd_X.nii.gz -div Tstd_Y.nii.gz –sqr palm_vox_rsqstat.nii.gz

This works fine when I don’t have any nuisance EVs, but when I include nuisance EVs my r-squared images have values outside the range [0 1], so I presume this is not a valid approach when nuisance EVs are included.

Is there a way to use voxelwise regressors in order to get fstat and rstat/rsqstat images?

I’ll copy below my ‘working’ command that generates the cope and tstat images. It is this command that I try to add –pearson and –f / -fonly flags to without success. Any help would be greatly appreciated. Thank you.

# Configuration file for PALM.
# Version alpha111, running in MATLAB 9.4.0.813654 (R2018a).
# 06-Sep-2018 22:46:48

-i Y.nii.gz
-d design.mat
-t design.con
-evperdat X.nii.gz 1 1
-m mask.nii.gz
-n 1
-demean
-noniiclass
-saveglm
-o output/palm

X.nii.gz and Y.nii.gz have 18 volumes each; design.mat has three columns, the first of which is the dummy EV, and the other two are nuisance EVs; design.con is [1 0 0]; when I add it, design.fts is [1].

########################################################################

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1