BOLD brain mask depicted in the ROIs reportlet does not correspond to the mask produced in the outputs

[chrisgorgo]

Background: https://neurostars.org/t/fmriprep-1-0-6-2-issues-semi-random-brain-mask-failure-and-super-bright-areas-in-epi/1303/8

EDIT OE: (moved to new issue #982)

reportlet:

output mask:

[chrisgorgo]

Seems like a 3dUnifize issue.

[chrisgorgo]

Similar problem is also visible in: https://openneuro.org/datasets/ds001216/versions/00002?app=FMRIPREP&version=48&job=5a7b6fd4831753000151c620

https://s3.amazonaws.com/openneuro.outputs/89832c01e43fee4b165e81a351af7b26/37c908e6-f88d-46e3-a0e3-d9c7e2dab4ac/fmriprep/sub-30009.html

[oesteban]

I can confirm that this is also happening on our Phase 2 test base. I'm seeing this on ds000237/sub-0[3,6] and some edge case on sub-01.

[oesteban]

Identified the source - this mask calculation is failing horribly:

https://github.com/poldracklab/fmriprep/blob/3e7d830e0322435532e875ec8c51213f9f6fec3f/fmriprep/workflows/bold/util.py#L176

[chrisgorgo]

This might explain #964, but why this mask calculation fail is not visible in the ROIs reportlet?

[chrisgorgo]

I don't understand why you renamed this issue. I believe the original problem is still unresolved - the mask plotted in the report comes from a different estimation process than the one (two?) leading to _brainmask file in the outputs.

[oesteban]

Sorry for the name change. I didn't understand the issue.

I've opened and closed #982, which is the issue that has been really addressed, restored the name and reopened this issue.

Also, I hope this message triggers a notification that makes all aware of these changes.

[chrisgorgo]

No worries. Thanks for clarifying.

[chrisgorgo]

@oesteban #1002 mentioned that the fix was only for the MNI part - if the T1w requires fixing maybe we should reopen?

[oesteban]

T1 was also fixed (a726ac4)

[chrisgorgo]

Great - thanks! Best, Chris

…

On Thu, Feb 22, 2018 at 4:32 PM, Oscar Esteban ***@***.***> wrote: T1 was also fixed (a726ac4 <a726ac4> ) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#963 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAOkp1iPxN3f1gft3jkA0jEctzzmTvRPks5tXgcWgaJpZM4R-qOq> .