top of brain cut after preprocessing using version 0.22.0
jhauneuro opened this issue · comments
Summary
The top of the brain is cut off after preprocessing using 0.22.0. The same dataset had full brain coverage after preprocessing using 0.19.1. Were there changes made to the image field of view between qsiprep versions 0.19.1 and 0.22.0? What could remedy this?
Additional details
- QSIPrep version: 0.22.0
- Docker version: 25.0.4
What were you trying to do?
Preprocess my dwi data.
What did you expect to happen?
Successful preprocessing.
What actually happened?
Preprocessed data cut off by several slices at top of the brain. Brain position shifts up between before and after preprocessing.
Reproducing the bug
Screenshots before preprocessing with 0.22.0:
and after preprocessing with 0.22.0:
Screenshots before preprocessing with 0.19.1:
and after preprocessing with 0.19.1:
I've experienced this too. It looks like a bounding box that crops tight along the top and bottom of the brain - unfortunately for some subjects as you've noticed it crops off the top of the brain (including projection fibers). I'd also like the cerebellum to be left intact - and the bounding box crops that in half too.
@jhauneuro and @johnaeanderson do you see cropping in the preprocessed images too? Or is it just in the figure?
Hi @mattcieslak, the cropping is definitely in the final preprocessed images as well as the figures.
Below, I'm using mango to overlay the file sub-201_rec-NORM_space-T1w_desc-preproc_dwi.nii.gz on my preprocessed T1 weighted image sub-201_desc-preproc_T1w.nii. On the sagittal slice, we can clearly see the aggressive cropping.
Thanks for looking into this!
@mattcieslak it's cropped in the preprocessed images. In the 0.19.1 output a bit more of the cerebellum is cropped (and the top of the brain is preserved) whereas in the 0.22.0 output the top of the brain is noticeably cropped - the cerebellum is slightly cropped too but this is also the case in the raw dwi data. I'd prefer to not lose the cerebrum at the risk of more cropping in the cerebellum if that is possible, or ideally no brain cropping at all. Thanks!
Circling back to this as I have now encountered this issue in multiple different datasets including in-house data and data from ABIDE-II in multiple subjects. Reverting to qsiprep 0.19.1 would not be ideal, however, it does provide results with better brain coverage/less brain cropping.