DWI to Tissue Probability Map Registration

, ,
1

Hello,

I need to register my DWI to SPM12’s MNI152 (which is IXI549). Their template, called TPM.nii, found at spm/tpm/TPM.nii, is their version of MNI152. It is 121x145x121x6, with a voxel size of 1.5mm^3. The 6 is the 6 different tissue types in this tissue probability map. My DWI dataset is 140x140x140 with a voxel size of 1.7mm^3.

I need to register my DWI to this TPM, so that I can apply the found warp field to my streamlines, thus warping my streamlines to SPM’s template. I can’t find anything about warping from DWI to TPMs. Has anyone ever tried this? Are there any tools to do this?

My current idea is to:

  • create TPM’s from my DWI
  • parse out the WM, GM, and CSF, components from my TPM-DWI
  • Register each tissue type to the SPM-TPM tissue map
  • sum these warp fields
  • apply this to my streamlines.

Best,
Hank

2

Hi @hankghgold,

Warping DWI is generality not as good as warping something else to DWI. When you warp DWI you have to rotate the b-matrix accordingly, which can be a pain. Why did the tutorial I linked to your earlier post not work ( DIPY — dipy 1.8.0 documentation )? I also do not know how you would align a DWI to a TPM since the image contrasts are not going to be the same. You would likely need a T1 or similar image in each space, perform that registration, and then apply it to the DWI or TPM to align them.

Best,
Steven

3

Seconding everything @Steven said. Just to add: one trick you can use to make a “T1w-like” image in the DWI space, if you don’t have that, as a target for registration with the T1w MNI template (in which space that TPM image is defined) is to generate an anistropic power map of the DWI data. This idea was originally proposed by Flavio Dell’Acqua a few years ago (in this ISMRM abstract: (ISMRM 2014) Anisotropic Power Maps: A diffusion contrast to reveal low anisotropy tissues from HARDI data.). We also had a preprint that explains this image contrast and some of its benefits: https://www.biorxiv.org/content/10.1101/864108v2, and there is code to calculate this image in DIPY. An example is shown here: DIPY — dipy 1.8.0 documentation

4

Hello,

Thank you guys both for the responses. My issue is that SPM uses a custom MNI152 template, that seems to be slightly different from the standard MNI152 space. My lab uses SPM for fMRI analysis - and we now need to map our dMRI structural connectome to this same space.

We are planning on using dipy’s streamline warping algorithm to apply the warp not to the dMRI itself but to the streamline image. Apologies if that was unclear.

I’ve used that anistropic power map technique successfully with T1 registration, but I have been unsuccessful so far in trying to source techniques to create a t1-like image from TPMs.

Best,
Hank