search

FCP-INDI / C-PAC

Configurable Pipeline for the Analysis of Connectomes
https://fcp-indi.github.io/
GNU Lesser General Public License v3.0
61 stars 39 forks source link

šŸ› Using the unfiltered head motion parameters (rotation and translation) for nuisance regression even for forks with the filter on? #1985

Closed shnizzedy closed 5 months ago

shnizzedy commented 11 months ago

Describe the bug

unfortunately it is looking like the output timeseries from the preprocessed data are identical for notch-none notch-notch4fc0p31bw0p12 and the run-specific notch pipelines only when there is no censoring. Based on this, I am wondering whether c-pac is using the filtered Jenkinson FD timeseries for censoring, but still using the unfiltered head motion parameters (rotation and translation) for nuisance regression even for forks with the filter on. ā€• https://danlabc-pac.slack.com/archives/C01AESSCX8W/p1692104120536609

Expected behavior

Filtered head motion parameters are used for nuisance regression

Acceptance criteria

Additional context

Itā€™s somewhat by design and mostly a mistake. When I separated out filtered-motion-parameters from motion-parameters so we could output both, I didnā€™t update the node that take motion-parameters as input to filtered-motion-parameters, which I assume they all should? Iā€™ll make that update

nimzodisaster commented 10 months ago

I am interested in using the filtered motion parameters. If I pull the nightly, will it have the latest patch for this, or is that still in development? Also, am I right in understanding that if I JUST choose filtered, it will be produced correctly in 1.8.5?

thanks

shnizzedy commented 10 months ago

@nimzodisaster

If I pull the nightly, will it have the latest patch for this, or is that still in development?

If you want a use-at-your-own-risk in-development version of this feature, you probably want ghcr.io/fcp-indi/c-pac:filt_downstream (that's the branch image for #1989 which is the PR to address this issue)

Also, am I right in understanding that if I JUST choose filtered, it will be produced correctly in 1.8.5?

Yes, it should be, except FDJ won't reflect the filtering in <1.8.6! This issue arose in the fix for that one, both in the 1.8.6 development cycle

shnizzedy commented 5 months ago

Fixed in 1.8.6