Hi All,

Can I use mrtransform to do color FA warp. like fod reorientation?

best,

Zifei

Hi All,

Can I use mrtransform to do color FA warp. like fod reorientation?

best,

Zifei

`mrtransform`

is designed to reorient ODF images. If your **really** wanted to, you could convert the tensor eigenvectors to the fixel format and reorient those using `fixelreorient`

.

I thought about it as well. But I have no idea how to make it.

I saw the command fixel2voxel and voxel2fixel.

Previously I try to replace split_data or dec_scaled data to the color FA, but doesn’t succeed.

Could you say more details about it?

Thanks,

An important clarification on this one:

When any orientation information is encoded as RGB, this involves taking an *absolute value* along each of the three axes. This can be thought of as projecting any direction on the sphere into the *positive octant*: the 1/8th of the sphere where x, y and z components of the direction are all positive.

Now imagine that you have four voxels, each containing a single fixel:

- Fixel is in +x,+y,+z direction
- Fixel is in +x,-y,+z direction
- Fixel is in -x,+y,+z direction
- Fixel is in -x,-y,+z direction

If all four of these directions are at exactly 45 degrees with respect to all three axes, then the RGB-encoded colour for each of these fixels will be *identical*; this applies to any RGB encoding of direction, including colour FA from the tensor model.

Now imagine that you wish to apply a reorientation of these fixel directions. Clearly, as soon as you apply *any* rotation, these three fixels should no longer going to have exactly the same colour as one another, since the impact of that rotation on the projection of each fixel direction along the three spatial axes will differ. Yet, if *all you have* are the RGB values for each voxel, the data you have in each voxel is *identical*, and therefore there’s no information by which each voxel could possibly be treated differently.

This ambiguity arises because by taking the absolute values along each of the 3 spatial axes, you are losing the orientation information that is required to perform reorientation. It is necessary to reorient the *orientation information* (e.g. eigenvectors) *before* calculating the RGB projection thereof.

Rob