New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Contours in space #281

Closed
effigies opened this Issue May 31, 2018 · 10 comments

Comments

Projects
5 participants
@oesteban

This comment has been minimized.

Copy link
Contributor

oesteban commented May 31, 2018

Yep, I've been seeing those artifacts more often. Something must have changed in nilearn proper.

@chrisfilo

This comment has been minimized.

Copy link
Contributor

chrisfilo commented May 31, 2018

@effigies

This comment has been minimized.

Copy link
Collaborator Author

effigies commented May 31, 2018

Their last release was March 12.

@effigies

This comment has been minimized.

Copy link
Collaborator Author

effigies commented May 31, 2018

@chrisfilo No. We pin a specific scikit-learn, but not nilearn.

@oesteban

This comment has been minimized.

Copy link
Contributor

oesteban commented May 31, 2018

March 12 seems like a probable date for these changes. I have also seen this in mriqc plots.

@effigies

This comment has been minimized.

Copy link
Collaborator Author

effigies commented May 31, 2018

If we want to try to reproduce with nilearn==0.4.0, the above was ds000164/sub-001.

https://openneuro.org/datasets/ds000164/versions/00001?app=FMRIPREP&version=61&job=5b0d753f2afa3badacec83df

@SRSteinkamp

This comment has been minimized.

Copy link

SRSteinkamp commented Jun 18, 2018

Just to understand this from a users' perspective:
Are these visual artifacts (e.g. some issues in the plotting function) or are these artifacts indicative that something went wrong in the ROI creation / normaliztion etc.?

@effigies

This comment has been minimized.

Copy link
Collaborator Author

effigies commented Jun 18, 2018

Hi @SRSteinkamp, good question. Our understanding is that it's an issue with the plotting function (we've looked at some actual masks, and there's no indication of non-zero values there). In any event, we plan to look more closely at it and resolve the issue, whether there's a plotting problem in the upstream library or something subtly wrong with our usage.

@mpcoll

This comment has been minimized.

Copy link

mpcoll commented Jun 22, 2018

Hi, In case this helps, I also get this for all my participants for BOLD, T1 and field maps. The actual images seem to be fine upon visual inspection. I did not notice this when I was using a previous version of fmriprep but I can't remember which one.

I'm currently using version 1.1.1 in a docker container.

sub-001_t1w_seg_brainmask
sub-001_phasediff_fmap_mask
sub-001_task-mvpasogl_run-1_bold_rois

@oesteban

This comment has been minimized.

Copy link
Contributor

oesteban commented Dec 27, 2018

regardless the nilearn version, this could be easily fixed improving the bounding box calculation to exclude slices with few pixels in the mask.

For instance, in the latest examples from @mpcoll, we can see how the first row is mostly brain stem in the two first cuts, and empty or almost empty in the last cut. In other words, those three cuts are almost useless. Surprisingly, for the coronal view (last row) of the fieldmap magnitude and bold registration, it seems that the bounding box was maybe too tight.

EDIT: that leads me to think that maybe we are calculating the bounding box in ijk coordinates but nilearn is taking xyz anyways. This could be an orientation problem when generating the plots.

@oesteban oesteban transferred this issue from poldracklab/fmriprep Jan 18, 2019

@oesteban oesteban added this to To do in 0.5.x Jan 18, 2019

oesteban added a commit to oesteban/niworkflows that referenced this issue Feb 7, 2019

@oesteban oesteban moved this from To do to Needs review in 0.5.x Feb 7, 2019

@oesteban oesteban closed this in 7d50907 Feb 7, 2019

0.5.x automation moved this from Needs review to Done Feb 7, 2019

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment