Very bad bbregister fail after fMRIprep

Ran a total of 21 func runs over 3 days for each subject in a study, only 1 run for 1 subject seems to fail to register.

We’re running fmriprep 20.2.x LTS on an HPC, which has been robust across all other datasets we’ve piped through it. Here’s the bbregister image from the fmriprep HTML: Dropbox - preprocess_fail.gif - Simplify your life

All other func preproc scans (collected using same sequence etc.) look good, T1 origin is at AC, and visual check on image registration between native func image and others from this subject also looks good.

About to go through our fmriprep working directory to try to trace the error, but the log file doesn’t report anything about this scan in particular so I’m pretty stumped at this point to be honest. Any suggestions appreciated!

Im a bit new to fmriprep myself so this could be totally off, but is the T1 reference image for this subject good? At least in the gif, the background image looks all slanty and distorted. Alternatively, did this subject have a lot of issues with head motion?

T1 does look reasonable imo

There’s a couple huge motion spikes in there, definitely didn’t help and worth looking into the working directory stepwise files to see if that caused it! That said, this same person also had 1-2 big motion spikes in several of their other scans which did not lead to the same bbregister failure.

Gotcha, I’m not too sure myself then. Is this subject’s mean FD typical too for your dataset? I know its task dependent but a mean of 0.38 could be seen as high for some analysis (I know the default threshold for exclusion in MRIQC is 0.2mm IQMs for functional images — mriqc 23.1.0 documentation) but that definitely should be taken in the context of your data.

Hard to say what could have caused that sort of result. Did you look at the original BOLD file in detail? Consider using MRIQC to see if there are anomalies that are not immediately obvious.

Still not seeing anything after MRIQC, and looked at the BOLD and can’t seem to spot anything myself…

MRIQC reports and the offending func data here in case you might have any suggestions! Dropbox - preprocess_fail