NPAC Internal Weekly Notes

• Zoom
• Prev (20241107 - 20250820) tetra confluence

• undergrand BME for NM
  • Bea and Finn *
• bids2nda: E Lorenc (brown), Yarik (Dartmouth)
  • https://etherpad.opendev.org/p/bids2nda
  • case for python: bids library, nda-tools, uvx/pip install
• MRSI – chan or andrewR (Finn) *
  • phanom paired with sesssion? (inhomo confound)
  • rfpulse in empty spectrum. ERCT
• redcap ↔ xnat incidental finding
  • api - small scripts: connect, pull, push (id=0 + force=True is new)
  • need to review fields (instruments)
  • survey links. then XNAT
  • ppt to questions
  • Jira
• protocol QC
  • exclude seqno > 100
  • series number for same sequence name within session (color if before or after error; matches template)
  • xnat tagging
• MRSI: gam adjust;
  • line width (fwhm) / SNR from LCModel? use matching cs.coord FWHM=….
  • grep FWHM LCModel/v2idxfix/*/csi.coord * L/R Cad. Thal
  • Cr for gam adjusted
  • Samira access to rhea for coords?
  • https://pubs.rsna.org/doi/10.1148/radiol.2402050314
• neurodesk: containers for OpenRecon; BART
• mallory on rhea/npac@cerebro2. ask nathan oac_ix_luna

• Neuromelanin MP2RAGE
  • phillips using whole brain & not CC (most stable voxel in residual?) for reference
• brainlife.io
  • install VMs
  • create/revive xnat connector
  • build bids curator ?
  • port file-curator alert?
• new tSequenceFileName in DB; can parse exar1 files exported from scanner.
  • https://github.com/NPACore/mrrc-hdr-qa/commit/ff31712ea02b4abcb2fad52fa57a33026d65e861
  • https://gitlab.com/tobiasrautenkranz/exar1-read/

AP

• investigating SR edits to NM Toolbox, only two changes were made. Unfortunately, there is no QC to check. Neuromelanin
• Changes are addition of antsMultivariateTemplateConstruction2 (builds 'unbiased' population template) & custom masks were used
• Found QC for some runs of the NM Toolbox that we can look over on /eru.
• Apparently there are still some issues with the modified pipeline (bright spots, have not seed this). Deepak had a thought to try MATLAB computer vision techniques, similar to microscopy segmentation, to delineate the boundaries of the ROIs.
• How precise is estimation of <x,y,z> motion from MRI, and would a gyroscope really improve this? Or is the internal motion estimation sufficient but not optimized yet? Things to consider, warping due to magnetic field inhomogeneity, second order effect?

* motion comparison with undergrad. correlation with param and tools. reho, connectivty/def mode graph?

• see Ted S.; fronteirs paper

* gyroscop. needs grant

* oac /eru; scan data * TODO: send scripts/RA notes on eyetracking (Silk lab; BRIDGE. camera focus)

• WF
  • scan_data mounted!
  • Shim values bad for P3 X; P1 X,Y. DailyQC updates needed for uT/m2 calc

• EH
• broke apart the main() method in email_latest_flip
• created the tests for it as well
• next step is to move to cerebro

• AP downloaded flywheel cli, logged on successfully, reading docs
• Singularity image of neuromelanin pipeline seems to be referencing a relative path?

• WF: QA Design and progress fw-mrrcqa FID automation and weekly upoads ~ 8h

1. mount update

* EH

1. Started the refactoring process of the main method

• AP no updates; last week: programmers happy hour

1. gear for NM on FW. cerebro? not needed

• cerebro / mrqart
  • git clone
  • cron onto cerbro
• more tests – refactor main, no global dict
  • using template checker; refactor email
• Tetra: DA study reproduction

• SSL resolved; gyrus down;
• iFFT rotate kspace w/ CHM code; write to nifti broken ~ 2 h
• pscyhopy task for grasping. Alex found EPrime task instead? ~ 3h
• QA IT Page matlab plotting: revise for R2019a. fix temperature pull ~ 2h
  1. TODO: file-curate for FID | .5h

* MRSI - Schefer ROIs

* NM docker works; bash script tested on oac-moon, testing singularity on CRC (Friday)

• Convert to Flywheel gear

• MRSI - shaefer ROI, CRLB,
  • sameria has list of PIs/projects
• NM docker works; bash script to run, testing singularity
  • on /eru
  • https://fw.mrrc.upmc.edu/
  • fw gear: manifest.json (input spec, docker container info); run.py to use flywheel config spec in manifest; CLI
• access to WPC-8986@cortex (10.48.88.14) for /disk/mace2/scan_data/WPC-8986/ rsync
• /etc/host edit set xnat.mrrc.upmc.edu to external 157.229.38.7

api_key

1. FWHM for fw-mrrcqa diff ~ 8h!
2. duo-htop (+openconnect for pitt) new bug ~ 2h worth it?
3. LLM agents thoughts to Alex ~2h(!)
   • Want case studies, prompt/org suggestions; time tracking
4. Phantom QC caught NuMoM sequence issue? ~1h
5. tat2 to dR2star, hand off to UMN ~ 4h

1. Built docker container and singularity image
2. Can clean up and add comments or just be done.
3. What is the next project / are the next steps?

• matlab → docker (matlab-branch commit

results = compiler.build.standaloneApplication("../Program/dostat.m");
compiler.package.docker(results, 'ImageName', 'fwmrrcqa-mlbase')
system('docker image ls --format=json fwmrrcqa-mlbase|jq -r .Size') % 3.57GB

• TODO: cerebro2 access to andrew/ via npac@…

1. QA IT Page false positive. need to support multiple templates ? ~ 1h
2. snr fast tSNR, re-benchmark. matlab 40s, octave hour+!? work on matlab container ~ 3 h
   • diff - use mcc's requiredMCRProducts.txt for smaller run time?
3. Flywheel Infrastructure cert diff

    curl -Iv https://fw.mrrc.upmc.edu |& grep expire # expire date: Jan  4 23:59:59 2026 GMT
   /raidzeus/flywheel/certs/2026/setup_ssl.bash -> /raidzues/src/hpc-env/flywheel/setup_ssl.bash

MRQART daily header-QA email cleanup & signal reduction ~ 2h

1. cleaned up mismatch formatting (expected → got), incl. FA / TA deltas
2. cut down email noise while keeping marquee vs any-diff behavior
3. made red-day emails quicker to scan for protocol issues
   • diff

Automated nightly template rebuild & MIA fixes ~ 3h

1. rebuild template_by_count at runtime to avoid stale templates
2. split MIA reporting into onboarded-study vs non-onboarded study
3. removed false-positive MIAs from lagging template updates

Non-conformance aggregation & post-hoc filtering ~ 2h

1. grouped parameter mismatches with counts across acquisitions
2. capped per-sequence examples without losing useful signal
3. filtered out post-hoc series (>200) from compliance checks

• report item per study/subj/sequence instead of study/sequence
• counts like (number sequences matching template, number match non-conforming parameter / total number of sequeneName in study)

• move from rhea to npac@cerebro2? cron @ 6am?
• pullout white/black list, params-to-check into config file?
  • start removing from blacklist. want to check EPI and fmaps
• tests with examples
• multiple templates for single acq? (nav_setter)
  • don't overwrite existing templates?
  • eventual interface for templates

1. Built docker container and singularity image
2. Can clean up and add comments or just be done.
3. What is the next project / are the next steps?

• NM for all sequences at MRRC? cerebro2 or CRC slurm?
• NM QC pulled out to work w/o batch (–participant-label sub-001 still makes QC pdf?)
• MRSI QPASA headless mode
• Flywheel QA (zshim) → cron, integrate tsnr, wiki page update?

1. Refined MRQART daily email
   • removed footer noise, tightened message formatting, swapped to SequenceType logic | 3h
2. Added full blacklist + filtering rewrite
   • implemented prefix-based exclusion, verified against database-wide stats | 6h
3. Rebuilt template_by_count
   • validated MP2RAGE/ACE templates, resolved all “no template defined” cases | 5h
4. Next week
   • Add optional template-diff reporting (highlight changes over time) | 4–5h
     1. maybe more spacing; html
   • Automate nightly template rebuild + logging into run_mrqart_daily | 3h
   • Explore improving TA (scan length)/FA mismatch descriptions | 2–3h
   • Also looking for feedback to make the email less dense
5. master.xlsx lookup for physicist – eventually send to them. email per study.
6. git commit
7. 2026: talk to phys; shot of image

-  Brain^wpc-8796 / test / ABCD_T1w_setter (Series 6): (3/5 bad)
              expected / have
        iPAT: null / p2;
        got R: -0.0008 0.4671 -0.0594 -0.8822 /  0.04 -0.02 0.01
     

1. dokuwiki This Wiki's Infrastructure
   • 32bit support for ipv6 into upstream | 5h
   • login button ewi:/wiki-mrrc/lib/tpl/readthedokus/tpl_parts, tpl_mobileheader.html (diff) | 1h
2. QA IT Page QA reporting T1w DB corrupt!? | 2h
3. BIDS fw example | 1h

• QA IT Page fast tSNR ~ 6h
• mrschedule MRRC schedule DB fixed slots ~ 6h

• Neuromelanin NM_Pipeline built and (almost) running. Current error line 31: ./for_redistribution_files_only: Is a directory. This is for the ENTRYPOINT command. Tried with and without ./ in ./for_redistribution_files_only/run_NM_Pipeline.sh
• TBO Task eye tracking at Bridge. (See email, waiting to hear from Helmet). Probably can try again with arguments to control executable?
• icaaroma Updates to ICA-AROMA? Can it be done in native space for precision psychiatry? Would this be useful? Hallquist BrainGnomes implements old ICA-AROMA pipeline already.
• Deepak's idea: Can MRRC make PDF's of scan parameters compatible with NDA upload spreadsheet format?
  1. https://github.com/WillForan/bids2nda
  2. Office of Data Management Adrian Sonheim (sleep)

• EH:
  • condensed the email
  • filtered using SequenceName → swapping to SequenceType
  • TODO:
    • listing allways flagged, finding SequenceType, blacklist | 5h
    • template sql code, run night? diff on the templates? | 1h + [10h]
  • (Bridge) FW job stuck for week+: rerun with logging (fw issue)
    • github repo for hafman code (xcpd scripts for extra atlases)
• AP:
  • Docker image built, can run, but some path errors remain. Very close!
  • EyeLink calibration working at Bridge. Grabbed code from Eyelink Testing scripts. Possibly the issue was the wrapper script. “O” in CAVAO sequence wasn't working. Ideas?
    • find path to control exe, WF TODO linux path
    • esc, ctrl-c; host pc
    • monitor: psychopy/pylink demo window(0 vs 1)
  • Triggy at some point
  • Differences in old vs. new pipeline for Deepak's TMS study (probably worth discussing)
    • ENACT (WPC-8986) ME fMRI AXCP (task) + resting;
    • See UNC IT slack; brainomes
    • fmriprep 25 Vs xcpd 0.5 vs 0.11; dlPFC cor. .3-.6 thres – was .9;; T1w wrong space for TMS software; use warren's for last subject
  • ICA aroma
    • validating
    • vs fmriprep cosine/regessors (CompCor regressors)
• WF
  • SPM DWI BIDS PR#87 accepted
  • tSNR mask working. 10x? speedup
    • diff
  • Framework for Image Reconstruction Environments (FIRE)
    • working Prisma1↔ptx at scan time https://github.com/NPACore/fire-recontwix
    • PR for tests in python ismrmrd-server PR#14
    • TODO: m tilt and MRSI slice position injector
  • Cerebro2 as NIST800-171 environment okayed by MR council
• Adding git diff
  • git log -p
  • 
• protected main branch? PR workflow?

• EH:
  • email not sending. test with python oneline
  • condense
  • oneday, all projects
  • ABCD T1 vnav → two filtered (dervied) raw data.
• AP:
• Docker image is building, need to add .dockerignore file and/or move base docker file and subdirectories up a level to save space in the build (~25GB)
• Deepak had a question about the MRSI Figure 1 ROI –> MNI Space. There was an overlap of the LNCD data and Sarpal data that got misplaced when Warren left. Deepak understood Warren to be working on this with Will.
• WF TODO: 3dMean of MRSI ROI for heatmap (need paths to volume)
• Sarpal 7T preproc /eru/shares/sarpal/test/7T/bids_correct/output/sub-2326/ses-20181019/rest/rest.nii.gz

skullstrip only applied to first volume, all other volumes are empty! rest.nii.gz corrupt? can use rest.3/rest.nii.gz to restore * hallquist can run on moon for new/missing subjects. see logs of existing for command line inputs * Eye tracking testing at Bridge Thursday at 4p? * WF TODO: linux path to control SDK – sample experiments * Triggy at some point

• QA emails sent! switch to all projects, just yesterday. alt all okay email
  • eventually link to project template
• EyeTracking BRIDGE still an issue
  • see C:\Program Files (x86)\SR Research\EyeLink\SampleExperiments\ from the EyeLink SDK
• Dockerfile; docker build $PWD for include files only
• Deepak: MRSI 7T; spack setup w/module load; MP2RAGE UNICORT

• alerts - nice looking html interface! emailing for next week
• NM pipeline - containerize, add masks
• flywheel was down at MRRC “upgrade may have been stuck in pending”

• mrrc-hdr-qa/mrqart/email_latest_flip.py
  • loop over all new: acq2sqlite.find_acquisitions_since
• NM pipeline - matlab container, bash. Phillips billy jewel? helmet
• FW (CRC into RHEL 9), clpipe deprecating, braignomes in dev?
• Feb-April new 7T
• process, async standup

• eye tracking - didn't work; cal crash. 2nd monitor off.
• watch out: axis flip from scanner if FOV angle changed to extreme
• review mrqart - loop over scans_since, template_check, report conforming

• P2 body coil replaced; new zshim? rerun stats next week
  • can we pull (all EPI sequence z-shims) from DB?
  • how variable? more variable over any time?
• MRQART IT |
  • prev: script / days
  1. move to python package! uv init . (name mrqart instead of *hdr*) | 7
  2. use DB = acq2sqlite.DBQuery()
  3. docstrings of template_checker and acq2… if/when needed
  4. check with make .test.pytest
  5. pyproject.toml console-script ⇒ ~/.local/bin (install mrqart)

• MRQART IT | prev5?
  • email setup! sending via sendmail – add andrew to list
    1. in cron on rhea
  • cron to mrqart
  • which fields?
  • mass pull: Flip Angle for just bold - (standard, median→mode value) TR, TE, FOV, …; Protocol Name: t2pe
  • does email send to list?
  • email format: date, scan, what mismatches
  • loop over projects - filter by date
  • eventaull into database
  • RT motion plots, thresholds (andrew?) – send to code
• Neuromelanin | prev1-5?
  • MacOS vs Linux Runtime? Including dependent files
  • [Seq meeting this afternoon]
• reproin/BIDS push to FW pilot users ideas?
  1. preprocess alex outstanding dataset. follow up (may be BIDS)
• P2 Rx failure? Tx box unstable? [need to check NM data]
• Flywheel debrief!

• Interactive QC Plot - EH date subset; added scroll and toggle. (commit)
• MRQART IT cron db not running since cerebro→warewulf!
  • check on email code. test code / cli call that flags an issue | 5

• Z-Shim Value QC
  • DONE: email update. Makefile change problematic (2-3)
  • DONE: recalc thresh; 1-month 1sd of zshim (from csv P2) | 2+1
  • add to repo, docs, maybe setup cron | 1
    • cron: read dataframe from https: url via pandas
• MRQART IT summarize different parameters inside seq (flip angle). identify strange values. start with bold.
  • send an email example report | 6
• Neuromelanin
  • docker w/matlab runtime like r2prime | 2h; 1-5
    • matlab and docker (cf. podman) setup – use nucleus
  • example r2prime-prisma/Dockerfile

• 96G 4min tar over globus (OAC server to clevland)
• singularity + xcpd →
• file curator on fmriprep output trigger email.
  1. email server connection a problem from google cloud.
  2. cf. cron fw ls check and mail
• Z-Shim Value QC
  • email update | 1
  • DONE: interactive plot “parse DATE as date”
  • recalc thresh; 1-month 1sd of zshim (from csv P2) | 2+1
• Neuromelanin
  • docker w/matlab runtime like r2prime | 1-5
  • example data into repo? bad idea w/o skull strip? git-lfs – use mni warped to subject

• connect a la CRC
  1. https://windows.cloud.microsoft/ → pitt sso → Psych AVD → FastX

• FW:Mike Gross - NDA vs BIDS
• Flywheel Gears annotate/document flywheel, complete_test. From workshop: file rename instead of bids-curate
• MRQART IT summarize different parameters inside seq (flip angle). identify strange values. start with bold.
• send an email example report | 6

👋 Hi Dr. Moon & welcome Andrew!

• flywheel workshop: push for cloud only. ITK-snap app with launcher.
• tools/infrastucture
  • move to dokuwiki (?) from confluence (assuming EWI) tetra confluence
  • issue tracker / “story points”; docs-in-code (sphinix) integration
• email config pytest updates (EH)
• deid PR pending (WF);a Michelle's question still open (email)
• EWI pending
• NCANDA pinged about Triggering Task Start

• NPAC: prioritize.
  1. Phantom QC
     • Z-Shim Value QC
       • 🌟 email robustness: send to MR accounts (EH: works w/ pytest tests/)
       • 🌟 adjust P2 threshold after TX box replacement?
       • 🌟 interactive plot Interactive QC Plot
       • rewrite with linux-inotify instead of flywheel
     • tsnr – rewrite for speed
       • matlab circshift (wrap around) vs ANTs alignment (cut off FOV)
     • FID channels (kspace signal)
     • MRIQC
  2. BIDS
     • 🌟 reproin push
     • 🌟 annotate/document flywheel, complete_test. From workshop: file rename instead of bids-curate
     • lncdtools bids gear
  3. dockerize lncdtools (in progress)
     • flywheel gear for bids tools
     • flywheel gear for tat2
  4. NM (AP/Deepak)
     • 🌟 dockerize + flywheel-ize
     • kernel density out of matlab (into AFNI 3dROIstat). also useful for tsnr
     • rewrite in shell/python
     • Jones: value into sheet, reverse, code to shift image (ADEPT study)
  5. deid.py
     • 🌟 demo/docs
  6. bids2nda - document, push to users
  7. MRQART
     • add motion
     • protocol QA in cron/DB build
     • push groups to use?
     • different users using different params: BOLD
       1. 🌟 summarize different parameters inside seq (flip angle). identify strange values. start with bold.
  8. Yarra @ NYU, overnight worker - TWIX transfer and processing. kspace recon. future proofing . offline high-load computation.
     • Yarra on “ptx” computer; high accel MPRAGE
     • OpenRecon – “inside” scanner
     • FIRE Siemens WIP - recon + fmriprep. docker. run external or internal
  9. Cerebro2 HPC maintaince: update from debian12 to 13 (bookworm→trixie)
  10. wiki conviences
      • added collapse button
      • keybindings to prose editor
      • php-scripts for pulling out log (could be plugins): IT Log
      • ☁️integrate with git-bug
      • ☁️integrate with doc generators
  11. ☁️ doxygen + sphinix: static site generator for scripts comments; monorepo
  12. ☁️ ML + MR sequence development
      • Deep Learning for reconstruction (kspace). better noise, faster sequence
        • need math background; vs. MR signal model for reconstruction. snapshot vs cumulative time.
        • eigen vector fit to remove noise: data + model to clean
        • optimization over large datasize (slow) vs. train and apply to inc speed (order of mag)
          • increase resolution
        • optimize code for GPU/CPU resources. CUDA/tensor/pytourch vs matlab optimizer
      • ML pulse sequence programming. RF, gradient, acq signal encoding (amplitude, phase)
        • EPI: train on echo time, matrix size, … parameters

- pitt email to andrew for slack;

• delete workign directroy free lots of data
• xcpd - shen atlas (empty files), rerun (workdir) populates tsv.

workdir=$(mktemp -d /tmp/XXXX)

• https://github.com/PennLINC/xcp_d
• tar search for dicoms
• https://github.com/ajeetdsouza/zoxide z-jump to dirctory on partial match of name
• shim_notify_test.py – test, then upload for all
• interactive plot Interactive QC Plot – 2 months
• autodocs for file-curate
• Enigma upload dicoms, waiting for BEAM dcm (maybe on flywheel)