External Email - Use Caution Hi Bruce,
Thanks for your response. I figured it out, but it took a while to troubleshoot. There were a number of issues at play, some due to my stupidity and some that may suggest feature improvements. 1. The primary issue was that the FOV of the T2 images did not cover the entire brain volume, presumably since the primary purpose of the acquisition was to inspect the ventricles and/or subcortical structures. 2. In this scenario, the current stable release appears to exit with the "Segmentation fault" error as described in my original message. The current dev version spits out the "MRIhistogramLabelRegion: constant image" error and then just hangs indefinitely with 100% CPU load. 3. It may be a good idea to implement a check at the beginning of recon-all (if a T2 or FLAIR flag is passed) to verify if the image contains the entire brain volume. If it doesn't, an error message could be raised, so the user has an opportunity to correct their mistake. 4. Once I realized the issue, I exported a different image sequence that contained the full brain volume. I then ran -autorecon3 with the new T2 image, but the process still failed. This is where I got stuck for HOURS. The reason for the fault was that the transforms for the T2 image from the previous T2 image are retained in the mri/transforms folder. Therefore, the transforms of the incomplete images were applied to the complete images, resulting in cropping of the brain volume to the point where it was then incomplete again. This ultimately led to the same problem. The mri/orig/T2raw.mgz image was in good shape, but the mri/T2.mgz would already exhibit the cropping. 5. It may be a good idea to implement a check that verifies whether the T2 image has changed and if the existing transforms correspond to the specified T2 image, possibly via a checksum that is stored in the T2raw.auto.* <http://secure-web.cisco.com/1Ca-6WOM9invcqQ6HF3AL5StykkRxAUrVlz_V5s9CLy8GOubgwdB9TcpdevemyJiWkkcR4P4Mp-7pY9cknQS2q9vslNih_ITBRt_Ax94GO74LurlNqX_EaXNWHqM30QSO_m12YU8QA1hJf7xhe_L-_oSwFYjD04a2KQ9bwzOMgXtqN15gh_JUi0QXWrrM1gQAsuT33ZOS4VhbBxiTIQZY0lA3_3JBFDKlwBFZzgzBLphzWgihSaFB7AA-5w7xLRAB4YOa3O1-m-fZtGP0ocgeQy8QkFv3v2CywRiWAtAI4OJpndqnP3iNuwfLB_fvU-mp/http%3A%2F%2Ft2raw.auto*/> and/or the T2raw.lta file. If it is determined that the input image has been updated and does not correspond to the transforms, the automatically generated transforms could be overwritten, thereby avoiding this issue. Or you could just recreate the automatic transforms regardless, since it seems that it would only be useful to retain transforms if they contain manual or external registration data. 6. If access to my images or logs would be helpful to you in the (potential) implementation of these suggestions, please let me know and I can upload them. I retained a copy of one of the problematic folders. There were a few issues I (may or may not have) encountered along the way. Unfortunately, I didn't keep detailed enough notes throughout my troubleshooting process and don't have the (compute) time to recreate these potential problem areas. It may be a good idea to verify if the following items cause problems: 1. I originally called recon-all with the -T2 flag, but did not include -T2pial. I don't recall if this created an issue or if the -T2pial flag is implied when passing a T2 image. Presumably, this behavior would be analogous for FLAIR images. 2. I worked with the -parallel -openmp N flag for a while and then suspected that this may have been part of the problem. I believe to remember that when using the -parallel flag, all of my T2 processing failed, while only the ones with incomplete image volumes failed when not using the -parallel flag. 3. I have one image series with 512 x 512 x 170 px dimensions and a FOV of 260 degrees. FreeSurfer is not happy about that. I tried to use -hires and -cw256 (which appear to be mutually exclusive). I eventually ended up simply cropping it with MRIcroGL. It would be nice to have an automatic feature in FreeSurfer that crops excessive FOVs, so that this glitch doesn't occur and this manual step wouldn't be necessary. It would also be helpful to have the option in mri_convert to specify both the crop size and the center of a crop. Currently, these options appear to be mutually exclusive. 4. Most of my T1 images are 256 x 256 px in size, but the T2 images are all higher resolution, either 384 x 384 or 512 x 512, but with a very large slice thickness of more than 4 mm. I don't know if it's possible and how it's currently implemented, but it would be nice to be able to process the T1 images in their original resolution to save on compute time, but then interpolate and refine with the higher resolution T2 images during the -autorecon3 processes. It seems that this would combine the best of both worlds. I hope this feedback is helpful. Best, Thomas *Thomas Kaufmann*, Dipl.-Phys., M.S. Pronouns: he/him/his Physicist | Audio Engineer | Disability Rights Activist *Arizona State University | College of Health Solutions* Speech and Hearing Science (Auditory & Language Neuroscience) 975 S Myrtle Ave, Tempe, AZ 85281 | Lattie F. Coor Hall Date: Wed, 9 Oct 2024 17:42:22 +0000 From: "Fischl, Bruce R.,PHD" <bfis...@mgh.harvard.edu> Subject: Re: [Freesurfer] Error During T2/FLAIR Pial Surface Refinement To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu> Message-ID: < dm5pr04mb126171d02f1f51dffe60c096ef...@dm5pr04mb1261.namprd04.prod.outlook.com > Content-Type: text/plain; charset="utf-8" Hi Thomas Can you check the input images? I've seen this happen when e.g. the skull stripping failed and the images are blank Cheers Bruce On Wed, Oct 09, 2024 at 10:37 AM, Thomas Kaufmann <tkaufm...@asu.edu> wrote: > Hello FreeSurfer Developers, > > I'm attempting to refine pial surfaces with T2/FLAIR images as described > here > (https://secure-web.cisco.com/1R4-8moFvOommy8UNXErfhOkkD_Bz4_9X2aJQpzXbNo9DgwIxW72SE9SZDo-enRjOr8lfrqJ0oEn4tFNNxKDbVQkYrF8sKH7h7tFMe_pyGmZHCwkUMcVXEWaGM8MmnSqATunbf6rffh63-56BO5FpFp-MjZwxd7pExvROAiE_jVNOePQQcqdjzr1KPoJLoEVG7lNLYHXgn8IpRGILsnAUcJ7OFvyBq7hpbL7ASPmPTJ6jtECsa-9AgAvzObWWqqAAojsjw4mJT0twQY3T-ze2sMQWjWO0L8U0AfZncf0SbRGjXFOzTzEz90YXQafGoyQfZ68KwX7vVuvawERMT-n_nQ/https%3A%2F%2Fsurfer.nmr.mgh.harvard.edu%2Ffswiki%2F > recon-all#UsingT2orFLAIRdatatoimprovepialsurfaces). For some of my > subjects I get the following error during the mris_place_surface subprocess > when I run recon-all: > > error: MRIhistogramLabelRegion: constant image > > error: MRIhistogramLabelRegion: constant image > > Segmentation fault > > The error occurs independent of whether I process T1 and T2 images > together with -all or if I run only -autorecon3 with the T2 or FLAIR images > added. > > I've searched the list and while similar errors have been reported, no > solutions have been proposed. Does anyone have any thoughts on how to > troubleshoot this one? > > 1) FreeSurfer version: > freesurfer-macOS-darwin_x86_64-7.4.1-20230614-7eb8460 > 2) Platform: macOS Sonoma 14.6.1 > 3) uname -a: Darwin MacBook-Air-M3 23.6.0 Darwin Kernel Version 23.6.0: Mon > Jul 29 21:14:04 PDT 2024; root:xnu-10063.141.2~1/RELEASE_ARM64_T8122 arm64 > 4) recon-all.log: relevant section below > > Thank you! > Thomas > > > mris_place_surface --adgws-in ../surf/autodet.gw.stats.lh.dat --seg > aseg.presurf.mgz --wm wm.mgz --threads 8 --invol brain.finalsurfs.mgz --lh > --i ../surf/lh.pial.T1 --o ../surf/lh.pial.T2 --pial --nsmooth 0 > --rip-label ../label/lh.cortex+hipamyg.label --pin-medial-wall > ../label/lh.cortex.label --white-surf ../surf/lh.white --aparc > ../label/lh.aparc.annot --repulse-surf ../surf/lh.white --mmvol T2.mgz T2 > > > Reading in input surface ../surf/lh.pial.T1 > > Not smoothing input surface > > Area 280714 0.42434 0.32508 0.000178 6.1090 > > Corner 842142 60.00000 30.99319 0.049953 179.7466 > > Edge 421071 1.04711 0.48130 0.006015 7.4374 > > Hinge 421071 18.87838 25.11022 0.000011 179.9979 > > Reading white surface coordinates from ../surf/lh.white > > Reading repulsion surface coordinates from ../surf/lh.white > > Reading in aparc ../label/lh.aparc.annot > > [DEBUG] CTABreadFromBinaryV2(): ct->nentries=36, num_entries_to_read=36 > > Reading in input volume brain.finalsurfs.mgz > > Reading in seg volume aseg.presurf.mgz > > Reading in wm volume wm.mgz > > MRIclipBrightWM(): nthresh=15378, wmmin=5, clip=110 > > MRIfindBrightNonWM(): 1814 bright non-wm voxels segmented. > > Masking bright non-wm for pial surface mid_gray = 67.5589 > > MRImask(): AllowDiffGeom = 1 > > MRImask(): AllowDiffGeom = 1 > > MRImask(): AllowDiffGeom = 1 > > Ripping frozen voxels > > Ripping vertices not in label ../label/lh.cortex+hipamyg.label > > MRISripNotLabel() ripped 6543/140359 vertices (133816 unripped) > > INFO: rip surface needed but not specified, so using input surface > > Ripping segs (eg, WMSA, BG, frozen) > > Starting MRISripSegs() d = (-2 2 0.5) segnos: 247 > > MRISripSegs(): -2 2 0.5 ripped 0 > > Reading in multimodal volume T2.mgz > > using multi modal weights > > vertex 70180: xyz = (-39.3637,-26.8713,28.9107) oxyz = > (-42.1603,-27.2597,31.4024) wxzy = (-42.1603,-27.2597,31.4024) pxyz = > (-39.3637,-26.8713,28.9107) > > CBVO Creating mask 140359 > > n_averages 4 > > Iteration 0 ========================================= > > n_averages=4, current_sigma=2 > > Computing pial target locations using multimodal (1) > > starting MRIScomputePialTargetLocationsMultiModal() > > max_outward_dist = 3, sample_dist = 0.1, pix_size = 1, whalf = 7 > > T2_min_inside = 110, T2_max_inside 300, T2_min_outside = 130, > T2_max_outside 300 > > inside_peak_pct = 0.1, 0.01, outside_peak_pct = 0.5, 0.5 > > wm_weight = 3, nlabels=0, contrast_type=1 > > Changed 198177 aseg cortex voxels to 0 > > Creating lowres distance volumes t=0.0164299 > > Creating white distance volumes t=0.307498 > > (box.dx, box.dy, box.dz) = (132, 242, 326) > > (region->dx, region->dy, region->dz) = (132, 242, 326) > > (region->dx, region->dy, region->dz) = (132, 242, 326) > > Creating pial distance volumes t=0.609579 > > (box.dx, box.dy, box.dz) = (144, 256, 336) > > (region->dx, region->dy, region->dz) = (144, 256, 336) > > (region->dx, region->dy, region->dz) = (144, 256, 336) > > locating cortical regions not in interior range [110.0 --> 300.0], and not > in exterior range [130.0 --> 300.0] > > t = 0.984043 > > Starting loop over 140359 vertices > > vno = 0, t = 0.984043 > > error: MRIhistogramLabelRegion: constant image > > error: MRIhistogramLabelRegion: constant image > > Segmentation fault > >
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Mass General Brigham Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline <https://www.massgeneralbrigham.org/complianceline> . Please note that this e-mail is not secure (encrypted). If you do not wish to continue communication over unencrypted e-mail, please notify the sender of this message immediately. Continuing to send or respond to e-mail after receiving this message means you understand and accept this risk and wish to continue to communicate over unencrypted e-mail.
