Hi -

I'm using PyMVPA to run an information spotlight on a nifti file I created with AFNI. The spotlight produces something akin to a dataset, which I then write out as a NIFTI file. The voxels in this dataset are floats containing the classification accuracy of a spotlight centered on that voxel; AFNI displays the results of this NIFTI file just fine.

The problem is that the information in the header is wrong, presumably due to the Frankensteinian method I used to build the dataset on which the spotlight was run, and perhaps the method I used to write the spotlight results into an .nii file, probably the latter, as 3dROIstats gives this message when I run it on the spotlight results:

** WARNING: NIfTI file tp101.novnum.nii dimensions altered since AFNI extension was added

The value of each voxel is a float from 0 to .91 or so; 3dROIstats -minmax reports this correctly. But the overlay text in the GUI is reporting a range from -96.91 to 109.91. Normally I wouldn't care about this, except the color scaling then colors all voxels uniformly, and I'd like them to be colored distinctly to indicate the different classifier results. Is there an AFNI command that basically says "reconstruct the header values for this file based on what you find in the dataset"?

POSTSCRIPT: I found out that I can get around this by 3dcopy ing the resultant .nii file to a HEAD/BRIK dataset; but I'm posting this anyway to see if there's a more direct route to fixing the .nii.