I think that would be a good encoding for single direction diffusion
data sets.

However, since NIfTI-1 can handle multiple dimensional data, it might be
more flexible (in allowing acquisitions with many different gradient
directions & amplitudes to be stored within the same NIfTI-1 file) if this
information could be included as a standard header extension instead?

Regards,
Guy Williams

On Fri, 12 Aug 2005, John Ashburner wrote:

> Dear NIfTI and diffusion people,
> I have had a few requests from people working with diffusion data who are
> interested in storing a direction vector in NIFTI headers.  I was wondering
> if there would be any objection to including the following new intent_code.
>
>
> #define NIFTI_INTENT_DIRECTIONAL    1012
>
> /* ! Data associated with a direction within the image volume.
>      Directions are defined with respect to the real-world
>      coordinates defined by the qform of the header (METHOD 2).
>      If qform_code is zero, then directions are with respect to
>      the sform (METHOD 3). If both sform_code and qform_code
>      are zero, then METHOD 1 is used.
>      intent_p1 = x component
>      intent_p2 = y component
>      intent_p3 = z component
>
>      p1^2+p2^2+p3^2 need not equal one (allowing additional
>      magnitude information to also be stored) */
>
>
> In this proposal, directions are associated with xforms rather than being
> native in the hope of simplifying rotating the data.  I hope this seems
> reasonable.
>
> For those of you on the diffusion mailing list who are confused about what
> this NIFTI stuff is, then the following may be helpful:
>     http://nifti.nimh.nih.gov/nifti-1
>
> All the best,
> -John
>