wiki

FSL Common Errors

Registration Problems

The image comes out twisted and flipped, and everything about the analysis seems correct, other subjects don’t have this problem Did you do registration with the standard space AND a high resolution, subject specific image? Sometimes, for a reason that I don’t understand, this happens, and by removing the middle registration image (the T2 high res) the problem is resolved.

Third Level FEAT

In Higher-level-stats log of a level 3 analysis “ERROR (nifti_image_read): failed to find header file for ‘filtered_func_data’ ERROR: nifti_image_open(filtered_func_data): bad header info Error: failed to open file filtered_func_data Error:: FslGetDim: Null pointer passed for FSLIO”
This error indicates that it had trouble locating a participant’s COPE, which either means that level 2 analysis didn’t work for a subject or two, there is a path error to a .feat directory, or some other reason a contrast for a particular subject cannot be found.
If you ever see an error in the main log or a particular cope’s log about “dof cannot be negative or 0” this means that you have zstats that are “NAN” (not a number) which means that you have errors in one or more of your first level FEATs. To find them, open up the filtered_func_data as an overlay in showsrs2 and look for huge spikes - the number that the spike occurs at is the subject that has the error. You can see an example at here Look in that subject’s first level FEAT in the log for errors that say something to the effect of:

/usr/local/fsl/bin/fslmaths filtered_func_data -Tmean mean_func 
++ WARNING: nifti_read_buffer(filtered_func_data.nii.gz): 
   data bytes needed = 557056 
   data bytes input  = 468775 
   number missing    = 88281 (set to 0) 

and

/usr/local/fsl/bin/fslmaths stats/zstat1 -mas mask thresh_zstat1 

echo 54628 > thresh_zstat1.vol 
zstat1: DLH=nan VOLUME=54628 RESELS=nan

and you will see an odd drop in the timeseries plots of the zstats. To fix this, try re-running and troubleshooting that particular subject’s group FEAT. Worst case scenario you can drop that person from the group analysis.