NITRC mri_reface Forum: help

RE: Resource recommendations

Paul Wright — Wed, 29 Apr 2026 15:19:45 GMT

Useful info, thank you. When I have more info after running an array of jobs I'll update the thread.
I agree there's probably an overhead to running in a singularity container.

RE: Resource recommendations

Christopher Schwarz — Wed, 29 Apr 2026 15:05:04 GMT

I agree that there's not much to gain in multithreading. If you watch a process graph through a run, most of the runtime is at 1.0 or less. In-development versions will have considerable performance improvements without multithreading, though.

Here's our stats for the past 6mo of defacing jobs on our cluster:

11374 jobs between 20251029 and 20260429.
MaxVMSize(GB)
Mean 6.23
Median 5.46
p95 21.05
p99 24.81

However, these are running from source code (not deployed/compiled) and not containerized, so it may be different for you. We're running a variety of image types and the higher usages are typically full-dynamic PET with 40+ volumes. For typical images with 1-5 volumes, I'd say 10G seems reasonable.

I think you really want the png snapshots for QC purposes, but they definitely increase requirements and dependencies.

Chris

Resource recommendations

Paul Wright — Wed, 29 Apr 2026 14:52:49 GMT

I am setting up an mri_reface pipeline on our SGE cluster. I've been experimenting with memory limits and multithreading. I'll share my results from my initial tests. If Chris et al. want to weigh in with any insider recommendations, I'd appreciate it, and I hope it may be useful to others.

I am using an apptainer image created from the mri_reface docker image.

Memory:

4 GB: stopped with MATLAB runtime out of memory error

8 GB: completed refacing but skipped generating the PNG snapshots

Generating renders for QC use
deNoseWrap failed. Skipping deNoseWrap.
#
# There is insufficient memory for the Java Runtime Environment to continue.
# Native memory allocation (malloc) failed to allocate 24 bytes for AllocateHeap

12 GB: ran with no memory errors or warnings

CPU

Single threaded:

OMP: Warning #96: Cannot form a team with 4 threads, using 1 instead.
OMP: Hint Consider unsetting KMP_DEVICE_THREAD_LIMIT (KMP_ALL_THREADS), KMP_TEAMS_THREAD_LIMIT, and OMP_THREAD_LIMIT (if any are set).

Using 4 threads (SGE submit script option `-pe smp 4`)

OMP: Warning #96: Cannot form a team with 20 threads, using 4 instead.
OMP: Hint Consider unsetting KMP_DEVICE_THREAD_LIMIT (KMP_ALL_THREADS), KMP_TEAMS_THREAD_LIMIT, and OMP_THREAD_LIMIT (if any are set).

I note the docs say there's not much gain to multithreading so I may ignore these warnings. I have not tested run time for different numbers of threads yet.

RE: Suffix "_T1w" not recognized

Christopher Schwarz — Mon, 27 Apr 2026 16:23:28 GMT

I added the BIDS suffixes in the last release after another request on this board. Actually what happened here is that you have two recognized strings ("T1w" and "MPRAGE") in the same filename, and I didn't code it to understand that some strings are aliases of the same type, i.e. not in conflict with each other. I tried to help people by adding more strings, and it bit me by not accounting for some inputs having >1 string (even though MPRAGE and _T1w together would be a pretty common name from dcm2niix). I agree that recognizing BIDS names is a worthy goal, but BIDS suffixes aren't detailed enough to handle everything (e.g. PET types you listed). Maybe the best answer is to prioritize BIDS suffixes when they exist and fall back to the rest when they don't. I think that would work, but I'll test some examples and think on it. That seems easier than building an "aliases" logic. Either way, thanks for reporting this. It's new to me, and I agree it's a bug. 
That said, if you're planning this at large scale, I'd suggest your wrapper-scripts make use of the -imType option rather than rely on the guessing. The taxonomy of image types and common names in the wild is just too complicated to rely on it entirely. PET tracers have multiple names. Sometimes MR sequences have different vendor names too. Most of the time we assume "FLAIR" means T2-FLAIR, but "T1 FLAIR" exist and should be de-faced like a T1. It's complex. I will try to improve things because your example is definitely a common case that should work, but it'll never be perfect. 
Chris

RE: Visualising changed voxels

Paul Wright — Mon, 27 Apr 2026 14:39:34 GMT

Originally posted by Christopher Schwarz:
<blockquote>
Hi Paul, We have plans to add a more full featured QC interface in a future release. The program is not really "mask"-based because there's some "as-needed" logic applied in the regions, but what we recommend is simply displaying abs(original.nii-deFaced.nii)>0 overlaid on the original in your favorite nii viewer (fsleyes for us), for inspection of "were brain voxels modified". You can use fslmaths or similar for that operation quite easily. This should always be in addition to also viewing the produced .png files for the "was the whole face replaced" question. 
Best regards, Chris Schwarz
</blockquote>
Thank you Chris. I had already done pretty much that for my test run, so glad we're thinking the same. I'll let you know what we learn testing mri_reface at scale and developing our QC approach.

RE: Visualising changed voxels

Christopher Schwarz — Thu, 23 Apr 2026 21:17:02 GMT

Hi Paul, We have plans to add a more full featured QC interface in a future release. The program is not really "mask"-based because there's some "as-needed" logic applied in the regions, but what we recommend is simply displaying abs(original.nii-deFaced.nii)>0 overlaid on the original in your favorite nii viewer (fsleyes for us), for inspection of "were brain voxels modified". You can use fslmaths or similar for that operation quite easily. This should always be in addition to also viewing the produced .png files for the "was the whole face replaced" question. 
Best regards, Chris Schwarz

Visualising changed voxels

Paul Wright — Mon, 20 Apr 2026 15:27:50 GMT

Is it possible to get mri_reface to save the masks it applied to the input image? This would make it easier in QC to identify cases where alignment failed and brain voxels were included.

Suffix "_T1w" not recognized

Paul Wright — Mon, 20 Apr 2026 15:25:34 GMT

I'm using BIDS naming, but the suffix "_T1w" was not recognized by mri_reface, so I had to specify "-imType" to run it. Error text:

"Unable to determine image type from filename for sub-paul_run-5_acq-MPRAGE_T1w. Use an input filename that contains one of: _T1, _T2, _T1w, _T2w,
_T2star, _T2starw, _FLAIR, _PD, _PDW, _FDG, _PIB, _AV45, _FBP, _FMT, _FBB, _TAU, _AV1451, _CT, or specify with -imType\n"

I note the help text lists different allowed strings:

"-imType <T1|T2|PD|T2ST|FLAIR|FDG|PIB|FBP|TAU|CT|AUTO>"

Please add BIDS suffixes to the code in the next update, and make the help text and error text match. These include:

T2star

T2starw

The suffixes below are listed in the error message, but not in the help text, neither in the BIDS glossary. I don't have an opinion about recognizing these, but please remove them from the error message if you decide not to recognize them.

AV45

AV1451

RE: /tmp/ directory issues on local

Anthony Kang — Thu, 06 Nov 2025 18:54:05 GMT

Dear Chris,
I just wanted to provide an update. I tried to do everything on WSL2 (mounting my drive, loading Docker, running the .sh) and, combined with the above findings in this thread, I was able to get mri_reface to run.
Thank you so much for your help!
Best,
Anthony Kang

RE: /tmp/ directory issues on local

Anthony Kang — Fri, 31 Oct 2025 13:59:53 GMT

Dear Chris,
Thank you so much for getting back to me so fast!
The part where I put the "C:/.../" is where I omitted my username, which only contains lowercase and uppercase letters.
Currently, how I tried to set this up is that I downloaded mri_reface 0.3.5 and Docker on my Windows machine. Since I couldn't figure out how to run the .sh file on Windows, I opened a git GUI window, ran the docker load command in the git window, and tried to run the .sh file from there, which then led to the error that I had in the post.
I wasn't sure if I could run mri_reface on my Windows PC because it is all that I currently have access to, so I just gave it a try. What kind of set up would be the easiest/would you recommend?
I greatly appreciate your time and consideration. My apologies as I am very new to this and that I am trying things that should not be done.
Best,
Anthony