Skull-stripping for BOLD image (with variable contrast)

, ,
1

Hello all,

I ran fmriprep (1.0.3) and some of the output data has poor/errors in skull-stripping (I think this is potentially due to issues with contrast in the image). I wondered if there was a way to try to troubleshoot this effectively (trying to use another skull strip template? Or if I could adjust the skull-stripping threshold via the docker command line?).

(run1 pngs)

run1_MNI152_no_mask
run1_MNI152_no_mask.jpg2624×998 655 KB

run1_MNI152_w_mask
run1_MNI152_w_mask.jpg2650×1014 637 KB

Another run from the same subject processed fine (run2 pngs)–

run2_MNI152_no_mask
run2_MNI152_no_mask.jpg2084×794 485 KB

run2_MNI152_w_mask
run2_MNI152_w_mask.jpg2072×788 392 KB

So I was unsure how best to proceed? Are there reasonable ways to reprocess the first run or should I simply exclude it? Any advice or thoughts are greatly appreciated!

Thanks much!
Jamie.

2

This issue should’ve been fixed a few versions ago. Could you try the latest version (1.0.12)?

3

Are you able to share the first 50 volumes of the failing time series? We’re working towards building regression tests for that portion of the pipeline.

2 Likes
4

Yes, I shared it via DropBox (pardon my delay in sending it along). It’s linked here:

1 Like