Hi All,

I have a question concerning obliquity, NIFTI, 3dresample and –deoblique options.

We have a particular scan that we’ve conducted on over 100 subjects now. At the end of last summer, we were encountering a problem with registration (using FSL), and decided to reorient all our subjects to be RPI (as this is only orientation that FSLVIEW reads in correctly, even though they say it should be LPI). We did not want to use fsl’s tools to do this as they are confusing so we used 3dresample –orient RPI on the raw data and output the result as .nii to be used in FSL. This all worked rather well, our registration issue was resolved, and we were entirely happy that we knew what was left and what was right.

More recently however, this has not being working. For this particular scan, we collect 39 3x3x3mm slices and so need to tilt the slices to fit the whole brain in – resulting in oblique datasets. So whenever we run the 3dresample command now and output as .nii, it doesn’t work – the data stay RAI. We can get around all of this by first deobliquing using 3dWarp (after timeshifting).

My question is what changed? I think that part of it may be that different people are running the scans and angling the slices more than before – so while the older scans were not considered oblique, the newer ones are. But is anything else different? And do I have to worry about all those older scans and analyses? We did check everything pretty thoroughly but still, this is beginning to keep me awake at night. For example, I was using 3drefit to do the deobliquing, but now I see a warning that was not there before: “Replace transformation matrix in header with cardinal matrix. This option DOES NOT deoblique the volume. To do so you should use 3dWarp -deoblique. This option is not to be used unless you really know what you're doing.”

I’m also wondering about the timeshifting pre/post deobliquing – I’ve tested this on a few subjects and sometimes it makes no difference at all, while other times it changes values very slightly, e.g., from 608 to 609. Does anyone have any insight into this? I ask this because some time ago and in a different dataset I encountered the failure-to-reorient problem in a handful of subjects and so I deobliqued them (using 3drefit!) before timeshifting. I thought I had researched this problem thoroughly but somehow I missed the instruction that timeshifting must be done beforehand.

One final question: given the problems associated with oblique datasets, is it better not to tilt the acquisition slices, or tilt them only slightly? We seem to get conflicting info on which is better… some people in our lab have been know to tilt the slices so much that they are almost vertical!

I really appreciate your help and advice,
Clare