[MINC-users] distorted tal transformations with 3T data.

[Dylan David Wagner]

Hey Minc Users,

      I've got 100+ mprage 3D MRIs from a Philips 3T scanner, all 
acquired with identical sequences. I'm attempting to transform them into 
tal space, but am finding the transformation is distorted for some, but 
not all, of the images.

      I've included an anonymyzed dataset on scratch1 (ddw_talsucks) in 
case anyone wants to poke around. The data were originally in analyze 
and were converted to minc using nii2mnc. To rule out conversion issues, 
I went back to the raw dicoms and converted some direct to afni and from 
there to minc using 3dAFNItoMINC (we don't have a direct dicom 2 minc 
script here). That didn't help.

       About 1/3rd of my images (30 out of 110) will not properly 
convert to tal space, checking both the original analyze header and the 
new minc header shows no difference in number of slices, step sizes etc. 
I even double checked the scanning protocols at the scanner and everyone 
is identical. The *only* difference I can find between files that work 
and those that don't, is that the working files have an intensity 
gradient (brighter in the middle) while the non-working files are more 
uniform. I know, you would expect it to be the other way around, but 
it's not.

       The range of intensities is pretty variable across all these 
files. Tried clamping some of the files with mincmath (0-300) before 
running mritotal, but that didn't help much. Nu_correct before mritotal 
on the bad files seems to help a little, but the output is still distorted.

       The dataset on scratch1 includes 2 "bad" files and 2 "good" 
files. You can see for yourself that the overall characterisitcs of the 
images are identical. But check out the resulting bad*_tal.mnc for some 
wild distortion!

       2 more things: we're using a SENSE coil to speed up acquisition, 
I'm asking around to see if a bad sense reference scan could be causing 
the variability between "bad" and "good" mris. And secondly, all these 
images normalize without fail and without distortion in SPM2.

		Thanks,
		       DDW