nipreps/nireports

Funky fMRIPlots with multiecho datasets

Closed this issue · 2 comments

oesteban commented

MRIQC/NiReports are executing the fMRIPlot in a very weird way:

sub-001_ses-20102030_task-rest_dir-RL_echo-3_desc-carpet_bold

oesteban commented

It seems this is not limited to ME-BOLD.

I believe this problem initiates in nipreps/mriqc#1167 -- the nireports and niworkflows implementations may be inconsistent.

oesteban commented

It doesn't look like the porting had any role. I pared down both implementations and there is no difference that justifies the new behavior. My theory now is that matplotlib figures are not correctly closed after saving, and the use of plt.gcf(). This, when run in parallel, seems to be the origin of corruption of figures. This would also explain #130.