This question was asked on the FSL list and I thought it would be pertinent on this list as well in relation to mri_convert.

---------- Forwarded message ----------
From: Saad Jbabdi <saad@fmrib.ox.ac.uk>

The bvecs are supposed to be relative to the image grid.

On 22 Sep 2009, at 13:34, Laura Mancini wrote:

Dear all,

In FSL, do the x, y, z components of the bvecs correspond to the image
axes (i.e. along the directions of the data matrix indices), or to the
patient axes as defined in the nifti header, or something else?

Many thanks,
Laura
------------------------------------------------------------

when one converts a dicom data set containing diffusion data with mri_convert it generates bvecs and bvals. the bvecs are unit vectors relative to some coordinate system. is this the coordinate system of the scanner or the coordinate system of the image volume?

cheers,

satra