Register DWI to T1 Before Structural Connectome?

connectomics

#1

Hi Everyone,

I hope this question has not been asked before.

I am creating structural connectomes based on the Human Connectome Project pipeline. I saw some discussions on registering diffusion images to structural (T1) images on the forum, and I was wondering if this is a necessary step to take at the beginning of the pipeline.

On the website, it says if no diffusion distortion correction applied, then non-linear registration is an option to consider. For a diffusion image, I am using the ones that have been denoised and eddy-corrected, and generating tractography based on anatomically constrained tractography framework.

I ran 40 people so far, and 3 of them are missing about 4 nodes. I was wondering why that is the case. Is there a possibility that some subjects are missing nodes because I am not registering dwi images to T1 in my pipeline?

Hope this makes sense, and thank you!

Melisa


#2

Hi Melisa,

I am creating structural connectomes based on the Human Connectome Project pipeline.

Presumably you’re not talking about using the HCP data itself, as in that case you’d simply use the minimally pre-processed data for which inter-modality registration has already taken place. If you’re trying to perform pre-processing of your own acquired data using the HCP pre-processing scripts, it was my belief that the post-eddy stage is responsible for registration between DWI and T1…

For any non-preprocessed data, registration between T1 and DWI really should be done. It’s not uncommon for participants to hear the MR scanner stop between protocols, and shimmy around to reduce pressure on uncomfortable areas of the head.

If nodes within the parcellation are “missing from the connectome”, meaning that not a single streamline terminates in / adjacent to that node, then that’s indicative of a gross alignment problem. Loading your DWI data in mrview, and then opening your T1 image using the “Overlay” tool in mrview should give you an idea of the magnitude of misalignment. Another approach is to feed your parcellation image through the labelcolour command, open that image in mrview, and also open the tractogram using the “Tractography” tool; that will show you the data that tck2connectome is operating on in its raw glory.

Rob