rick reynolds wrote:

> In any case, I'm not sure why you are expecting adwarp to work
> the same after a tlrc dataset gets changed from RAI to RAS
> orientation. Putting acpc and tlrc datasets in RAI orientation
> and having the transformations rely on it sounds like a solid
> design decision to me, not a bug. The slow part was that I was
> not sure enough to say that it must be RAI, up front. I only
> suspected and suggested as much.

Well, it is at least a documentation issue- the warp data is described as being given in DICOM coordinates (in README.attributes). RAS vs. RAI order effects the calculation of the 3D DICOM coordinates of any given voxel, but given those coordinates the transformation of the voxel should not depend on RAS vs. RAI. If this were the case, 3dresample wouldn't have to throw out the warp data in this case.

> So it seems that the clear way to go is using the afni-created
> tlrc dataset as the parent for warping the functional data.
> Please let me know if that does not work out for you.

That does indeed work. It drives me nuts, though- the fact that I clearly don't understand the management of the warp data means that if something breaks in our processing stream later on I won't be able to fix it.

-Joel

> - rick
>