Question

Question Posted 01/30/14:
When I download DTI images in NIFTI format, there are 46 NIFTIs that should include 5 B0 and 41 diffusion weighted images. However, I am wondering which images correspond to which directions. I know that the gradient table is available (http://adni.loni.usc.edu/support/experts-knowledge-base/question/?QID=97), but I have no clue which NIFTI file corresponds to which direction just by looking at the name of the file. And the header of a NIFTI file does not seem to provide any information about the gradient direction. Also, I know that it can be found in the protocols (http://adni.loni.usc.edu/methods/documents/mri-protocols/), but is there any way for me to find the b-value from the file?

Also, if I use the conversion tool, dcm2nii, the converted NIFTI files and the NIFTI files that are downloaded from ADNI have different quaternion values. The NIFTI files that are downloaded from ADNI have the following quaternion values: a=b=c=d=0. Is it part of ADNI conventions?

And could you please recommend me a good conversion tool that converts DICOM to NIFTI, if you know any?

Thank you very much.
Response posted 02/04/14 by Bret Borowski:
Thank you for your question to the ADNI MRI core.

After checking with some of the individuals responsible for DTI analysis of the ADNI data. Below are the responses to your inquiry.


46 gradients are
first five b0s + 41 directions in that linkage.
So you will have
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
1.000000 0.000000 0.000000
0.798000 0.603000 0.000000
-0.264000 0.553000 0.790000
0.930000 -0.353000 -0.099000
-0.159000 -0.779000 0.607000
-0.565000 -0.099000 0.819000
0.537000 0.448000 0.715000
-0.545000 0.818000 -0.183000
0.755000 -0.240000 0.610000
-0.168000 -0.985000 0.040000
0.226000 -0.958000 -0.178000
0.112000 -0.829000 -0.548000
-0.503000 -0.495000 0.709000
-0.491000 0.767000 0.414000
-0.717000 -0.126000 -0.685000
-0.527000 -0.848000 -0.058000
0.746000 -0.475000 -0.467000
0.587000 -0.289000 -0.756000
-0.223000 -0.916000 -0.334000
-0.827000 0.494000 -0.270000
0.599000 0.683000 0.418000
-0.477000 0.337000 -0.812000
0.934000 -0.046000 0.354000
0.229000 0.703000 0.673000
-0.150000 0.298000 -0.943000
0.151000 -0.970000 0.192000
0.888000 -0.109000 -0.448000
0.104000 0.425000 0.899000
-0.237000 0.830000 -0.504000
0.690000 0.611000 -0.388000
-0.105000 0.609000 -0.786000
-0.002000 0.002000 1.000000
-0.394000 -0.108000 -0.913000
-0.251000 -0.327000 0.911000
-0.807000 -0.245000 0.537000
-0.931000 -0.322000 0.171000
0.561000 -0.624000 0.544000
-0.711000 0.693000 0.120000
0.871000 0.365000 0.329000
-0.274000 0.186000 0.944000
0.398000 0.851000 -0.342000

when you open the nifti format image, you will see there are 46 volumes (the fourth dimension), so these 46 volumes are corresponding to the 46 vectors listed above.
Regarding the question "quaternion values: a=b=c=d=0",
In theory, dcm2nii won't change the quaternion values.


Internally, Mayo uses dcm2nii for converting DTI DICOM to NIfTI-1 format. The LONI interface to ADNI data creates NIfTI-1 formatted data which some users find problematic -- multiple 3D volumes instead of a 4D and sometimes people report weird (often anemic) geometry -- this happens for the fMRI data too. I don't know if the conversion is on-the-fly or cached. The only case I know of where the ADNI MR images are stored natively in NIfTI-1 is pre-processed T1 images. If you are really worried about the geometry in series other than the T1-weighted images it's probably best to get the DICOM and convert it yourself.