Synb0-disco

Development
1

Hi,

are there any plans to provide Synb0-disco as an alternative to topup, implemented in dwifslpreproc?

Best regards,
Stefan

4 Likes
2

I can generally strongly recommend Synb0-disco to anyone by the way: so far, this has performed extremely well on our data, allowing for excellent alignment with T1w images for ROI-based analyses and the like. Very impressive.

3

Hi Stefan,

I’m definitely a big fan of the concept behind Synb0-disco, but I’ve personally been so overwhelmed with other things that I’ve not even had the chance to try it yet, let alone decide how best to interface with it. I think allowing dwifslpreproc to accept a topup output path prefix (field coefficient image and movement parameter text file) as an input at the command-line probably makes most sense. dwifslpreproc is a bit of a monster, and I generally don’t enjoy going in there because there are so many possible branches of execution to consider; but I’ve now explicitly listed as an issue, and hopefully I will find the opportunity at some point.

Cheers
Rob

4

Hi Rob,

You can find a version of dwifslpreproc that supports Synb0-disco here: https://github.com/treanus/KUL_NeuroImaging_Tools/blob/development/kul_dwifslpreproc.

It’s alfa, but works for my data…

@ThijsDhollander, indeed works better than topup for my recent project on the dentato-rubro-thalamic tract. This goes from top of the brain to cerebellum, passing through certain brainstem nuclei. I have the strong impression synb0 does much better in the brainstem/cerebellum area.

Cheers,

Stefan.

1 Like
5

I overlooked the need for a FreeSurfer license file in there… that means that executing such within dwifslpreproc would be exceptionally clumsy if distributed to a wider audience. So passing the topup output prefix at the command-line is probably the most sensible interface-wise.

I really should get this in place in order to be able to process my own lab’s retrospective data, so I should try to have a go at this soon…

6

Yep, starting to see similar benefits. It seems to depend on the quality of the RPE data, and certain aspects of that quality, whether the one or the other strategy “wins” out. Definitely opened my eyes in terms of what’s useful to spend acquisition time on, and what not. In the meantime, we’ve got evidence from multiple labs; looking great so far!