ABIDE `fmriprep` vs. C-PAC

Hi everyone! Excuse my naivety as I’m quite new to the field! I am looking into the ABIDE I/II datasets. It seems that even though it was processed using fmriprep and the output is on their website, most people who use the data for deep learning autism prediction studies prefer the C-PAC pipeline approach.

For example in this paper the authors state:

We chose the data processed through the Connectome Computation System, without global signal regression but with band-pass filtering.

but no explanation as to why. Same with more articles describing their preference for C-PAC but no further evaluation or metrics. It would be interesting if anyone has insights on that.

With this in mind, what would be a good way to compare results between different pipelines? I am especially interested for the weighted degree centrality metrics. It’s difficult to create a ‘golden average’ with degree centrality. Is there any qualitative way to compare those?

Thanks a lot for any insight given :slight_smile:

Hello wizofe!

The Connectome Computation System is actually CCS, developed by Ting Xu et al. - though she is part of the same lab that produces C-PAC (of which I am a part of).

Here are some relevant links:

However, the C-PAC team is currently working on reproducing the CCS pipeline within C-PAC, to make use of its computational capabilities.

Please let me know if you have any further questions.


1 Like