I’m running a group-level analysis and have encountered an error that results in a core dump. The first-level analyses were run using the fsaverage surfaces, rather than self surfaces. I’ve already inspected the individual contrasts, loading the
fsaverage surface in tksurfer and rendering each subject’s contrast as an overlay, and would now like to compute the group-level results.
I first ran isxconcat-sess, and saw no errors indicated in the program output.
mri_glmfit was then called as followed, and produced the following output (same result for the rh):
$ mri_glmfit --y ces.nii.gz --wls cesvar.nii.gz --osgm --surf fsaverage lh --glmdir glm.wls --nii.gz
Reading source surface /home/chris/LDT/fsaverage/surf/lh.white
Number of vertices 163842
$Id: mri_glmfit.c,v 1.196.2.8 2012/11/01 18:51:41 greve Exp $
cwd /home/chris/LDT/RFX/LDT_fsaverage.sm4.rh/hi-v-lo
cmdline mri_glmfit --y ces.nii.gz --wls cesvar.nii.gz --osgm --surf fsaverage lh --glmdir glm.wls --nii.gz
y /home/chris/LDT/RFX/LDT_fsaverage.sm4.rh/hi-v-lo/ces.nii.gz
labelmask /home/chris/LDT/fsaverage/label/lh.cortex.label
Creating output directory glm.wls
Loading y from /home/chris/LDT/RFX/LDT_fsaverage.sm4.rh/hi-v-lo/ces.nii.gz
Saving design matrix to glm.wls/Xg.dat
Normalized matrix condition is 1
Found 149955 points in label.
ERROR: mri_reshape: number of elements cannot change
nv1 = 163842, nv1 = 537168
Pruning voxels by thr: 0.000000
Segmentation fault (core dumped)
This seems to have its root in a mismatch between the vertices in the fsaverage lh.cortex label and those in the contrast data. However, I carried out subject-level analyses in the fsaverage surface space specifically to facilitate a group-level
analysis in fsaverage surface space, so I can’t see where I went wrong. The only other post I could find that was similar was related to an attempt to carry out a group analysis in native surface space, which is not the case here.