Freesurfer bbregister warning using HCP data

Software Support
1

Summary of what happened:

Hi everyone! I am using the HCP S1200 3T dataset. When I try to run a simple bbregister to align my BOLD volume with my T1, I get this warning:

* WARNING: initial G-W contrast is negative, but expecting positive. If the mov data has a T1 contrast, re-run with --T1

Of note, my BOLD image does not look like a T1, but like a regular BOLD image - there is nothing about the signal or intensity that makes it look off or T1-like.

Also, I have already checked the recon-all wm and pial boundary files and they look really good.

Otherwise, bbregister runs with no other errors or issues. But, the final bbregister output is terrible:

Screenshot 2025-09-25 at 3.00.33 PM
Screenshot 2025-09-25 at 3.00.33 PM1018×1018 48.7 KB

So my questions are:

  1. Is this G-W warning something to take seriously?
  2. Has anyone had similar issues with bbregister using HCP data?
  3. Any other suggestions as to why bbregister is failing?

Command used (and if a helper script was used, a link to the helper script or the command generated):

bbregister --bold --s STRUCT --init-fsl --mov /project2/brunamar_1504/iMAP/iProc/mri_data_LR/100206/cross_session_maps/templates/100206_midvol_unwarp_2mm.nii.gz --reg /project2/brunamar_1504/iMAP/iProc/mri_data_LR/100206/cross_session_maps/templates/100206_allscans_meanBOLD_to_T1.dat --fslmat /project2/brunamar_1504/iMAP/iProc/mri_data_LR/100206/cross_session_maps/templates/100206_allscans_meanBOLD_to_T1.mat

Version:

Freesurfer 7.4.1

Environment (Docker, Singularity / Apptainer, custom installation):

High Performance Cluster SLURM environment with Freesurfer loaded as a module.

Data formatted according to a validatable standard? Please provide the output of the validator:

Data is in BIDS format.

Relevant log outputs (up to 20 lines):

Loading mov
ExcludeZeroVoxels 1
Reading in targ vol /project2/brunamar_1504/iMAP/iProc/fs/100206/STRUCT/mri/orig.mgz
Projecting LH Surfs
Loading lh.white surf
Loading lh.thickness for GM
GM Proj: 1 0.500000 2.000000
WM Proj: 0 0.500000 2.000000
Projecting RH Surfs
Loading rh.white surf
Loading rh.thickness for GM
Projecting RH Surfs
Using lh.cortex.label
Using rh.cortex.label
Computing relative cost
 0  -25.0 -25.0 -25.0   1.008519
 1  -25.0 -25.0  25.0   0.984354
 2  -25.0  25.0 -25.0   1.036814
 3  -25.0  25.0  25.0   1.065209
 4   25.0 -25.0 -25.0   1.029423
 5   25.0 -25.0  25.0   1.062277
 6   25.0  25.0 -25.0   1.030787
 7   25.0  25.0  25.0   0.981825
REL:  8  1.002380    8.199207  1.024901 rel = 0.978026 
Initial costs ----------------
Number of surface hits 3337
WM  Intensity  4812.0073 +/- 1677.8249
Ctx Intensity  4712.8356 +/- 1775.0430
Pct Contrast     -5.5893 +/-  36.3255
Cost   1.0024
RelCost   0.9780


WARNING: initial G-W contrast is negative, but expecting positive.
If the mov data has a T1 contrast, re-run with --T1

Screenshots / relevant information:

I have tried bbregister with other volumes from this same subject and I still get the same warning.

2

I found the solution to this, so for anyone with a similar issue: it seems like my initial registration using --init-fsl was not great, which led to BBR also not being great. I instead manually ran fslregister on my T1 and BOLD image and fed bbregister the output .dat file as an initial registration position, like so:

fslregister \
  --s $SESST \
  --mov ${MOVIMG} \
  --reg ${DAT_INIT} \
  --fsvol brainmask.mgz \
  --dof 6 \
  --niters 10 \
  --maxangle 30

bbregister --bold --s $SESST --mov ${MOVIMG} --reg ${VREG_REG} --fslmat ${VREG_MAT} --init-reg ${DAT_INIT}