Hi Yi-Fen,

Correct, by running dke you are not doing anything with tractography, just estimating the tensors and the tensor-derived parameters.

Ok, that makes sense, This is occurring from spm's dcm2nii functions, I believe. Dicom images are stored in LPS, but when they're converted to nifti the data is saved in LAS with information in the header (hdr.mat, hdr.private.mat, and hdr.private.mat0) which specifies a voxel to RAS (in mm space) transformation. I believe viewers like fsl use the RAS mm representation, which will look flipped in the A-P orientation and often rotated obliquely compared to what is in the dicoms (the rotation comes from the ImageOrientationPatient field in the dicom header, I believe, but can also be modified by other programs (like spm to set the anterior-posterior commissure coordinates ~ confusing right?!).

However, DKE should output a set of nifti images in the exact same space as all of the parameters, like b0.nii. These should overlay correctly in any nifti viewer and have the same general header information. If not, I am unsure what could be the problem. 

If you load these images into matlab; ie hdr = spm_vol(file_name.nii); img = spm_read_vols(hdr); the images should not look rotated obliquely, but they will likely still look flipped in the A-P direction compared to the dicoms; ie img = dicomread(file_name.dcm);

I hope that helps,

Russell