Converting to fsaverage from MSMSulc or MSMAll

[Nick Blauch]

Hello HCP experts,

I am working with HCP data, but prefer to use PyCortex for visualization, which I have set up with the fsaverage surface. Therefore, I want to convert HCP maps to fsaverage.

I am using the neuromaps toolbox to do this (leveraging wb_command), however, it appears to only work correctly with MSMAll surfaces, but not MSMSulc. Previously, I found the same issue with the hcp_utils toolbox when using their atlases. 

Using neuromaps, you can see an example below where the alignment is good for MSMAll, and is mostly good for MSMSulc, except for the presence of extreme negative values in certain areas. 

I am assuming this is because, as Emma Robinson points out in this blog,  each surface has its own sphere, and it appears these toolboxes are set up to use the MSMAll sphere.

 

How can I get MSMSulc data into the MSMAll space so that these other toolboxes will work? Within the subject MNINonlinear/fsaverage_LR32k folders, I see only one sphere, so I am a bit stumped. However, within the MNINonlinear/Native folder, I see one sphere for each of the surfaces (Native, MSMSulc, and MSMAll). Could I use these surfaces for converting group-level data between MSMSulc and MSMAll? 

 

Thanks very much in advance.

best,
Nick

[Tim Coalson]

MSMAll and MSMSulc use the same spheres, they just use different types of data for the registration.  The triangular patches imply that something bad happened while linearly scanning through the fs_LR standard spheres, but it is weird that there are occasional reasonable-looking values interspersed with the bad values (and some of the opposite, too).  It also appears that the same fs_LR vertices have bad values in each hemisphere (though the translation to a different mesh may have created some differences).

Without knowing what step introduced this problem, it is hard to help with it.  Have you asked the people who developed the neuromaps toolbox?

We have our own instructions for resampling between fs_LR and fsaverage, see #9 here:

https://wiki.humanconnectome.org/docs/HCP%20Users%20FAQ.html

We also have the colin flatmap surfaces in fs_LR, so you could display our fs_LR data without resampling it:

https://github.com/Washington-University/HCPpipelines/tree/master/global/templates/standard_mesh_atlases

Tim

--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hcp-users+...@humanconnectome.org.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/8b3d82e6-7ced-4092-a776-172df3c8d1e9n%40humanconnectome.org.

[Nick Blauch]

Tim - thank you very much for your reply. That makes sense to me (I had been confused as to why there would be separate spheres). 

It turns out this had nothing to do with registration, in line with your points. There was simply one subject with an apparently corrupted MSMSulc functional file containing many very high magnitude values, which propagated through the other steps of the pipeline. I was able to redownload the data file and fix the issue. 

best,

Nick