search

VisionandCognition / NHP-BIDS

Code for NHP-MRI (e.g., Retinotopy, Curve-tracing, Figure-ground) using NiPype and the BIDS format.
7 stars 1 forks source link

Allow combining runs with missing conditions/trials to modelfit #21

Open pcklink opened 4 years ago

pcklink commented 4 years ago

The fMRI (feat) analysis currently crashes when you try to include sessions that do not contain trials/events for all specified contrasts. It would be convenient if the workflow would detect this as an issue and automatically exclude either the contrast or the session altogether.

pcklink commented 4 years ago

@kanishkatks can you provide a log file with the precise error?

pcklink commented 4 years ago

@kanishkatks could you run a 'dummy-analysis' of this kind (1 good run CT without the central task, and 1 good run with the central task). It will throw an error and crash, but the logs will be informative and give us an idea where to start.

nhpatscnin commented 4 years ago

I have attached the slurm output in the following link. It didn't create a log file for some reason where I wanted it too. https://www.dropbox.com/sh/xwkzh5kyht43v4u/AAB4MGFrppPYo8X9c0u-JzuMa?dl=0

pcklink commented 4 years ago

Seems like this might be possible. Fsl should take 'empty' ev files if you code them as '0 0 0', which should yield an empty result image. For the second level we should than use the copes, not the low-level feat dirs. All untested....

pcklink commented 3 years ago

Putting this off. Doesn't seem necessary for now. Would only be helpful if one type of trial would be structurally ignored or something. Usually, it's better to fix this with training. The above solution might still work but is not worth the effort right now.