Tensor image looks strange under the OFD display

HI,dear developers,
I have created the diffusion tensor image with the following codes, and everything went successfully. But when I check the tensor image with mrview, it looks weird (see below).
What may cause this?

# create a mask
dwi2mask align.mif - | maskfilter - dilate dwi_mask.mif

# create diffusion tensor
dwi2tensor -mask dwi_mask.mif align.mif dt.mif

Note: align.mif is that conducted the DWI to TIw alignment.

Thankyou in advanced!

Best,
yuting.

1 Like

Hi Yuting,

If you’re talking about the black tensors, I suspect these correspond to tensors that happen to have been estimated with negative eigenvalues. Our fitting procedure doesn’t explicitly constraint the tensors to be positive definite, so this can happen. You would tend to see this in very anisotropic voxels due to noise, or indeed in voxels that are simply very noisy. This is why you can see many of them in the rim of CSF around the brain. The one you’re highlighting under the crosshairs is I suspect also in a noisy CSF region (presumably the Sylvian fissure?).

As to why it looks OK in the separate ODF display: that’ll probably be simply because the components of the tensor are interpolated from its nearest neighbours, so you won’t get the exact same coefficients as you would at the very centre of the voxel. You should (:crossed_fingers:) see the same thing if you untick the ‘interpolation’ check-box.

All the best,
Donald.

1 Like

Hi Donald,

Thank you for your reply.

As you said, this maybe caused by the noise, but I have denoised, why there still so much noise?

And according to your reply, this is just the display issue, it doesn’t seem to affect the metrics generated from diffusion tensor such as FA, am I right?

Best,
Yuting