Freesurfer

freesurfer@nmr.mgh.harvard.edu

9 participants
30092 discussions

QS for using cuda in FS6.0
by 巍魏 27 Feb '17

27 Feb '17

dear FS community： now i have many subjects may should rerun in FS6.0 so i thought could i use gpu for accelerating, and through searching in the mailing list i got some messages: 1. cuda was no longer supported 2. new features in cpu version were not add in gpu version as those above, is following example correct? in FS6.0 reconall_table i got this: mri_ca_register<https://surfer.nmr.mgh.harvard.edu/fswiki/mri_ca_register> -align-after -nobigventricles -mask brainmask.mgz -T transforms/talairach.lta norm.mgz $FREESURFER_HOME/average/RB_all_2016-05-10.vc700.gca transforms/talairach.m3z obviously it is new and different from mri_ca_register in FS5.3, so if i run mri_ca_register_cuda, it cannot get RB_all_2016-05-10.... and then may shutdown or run with the older version. the right gpu now was not setted so i cannot run a test, while i read a letter https://mail.nmr.mgh.harvard.edu/pipermail/freesurfer/2017-February/049923.…, that Dr Chris Adamson said gpu version ran sucessfully, which made me confused, dose gpu run the real FS6.0 with so many new features? hope some one can help best regards to you all! weiwei west china hospital ________________________________ weiqi77777(a)live.com

1 0

INTERSUBJECT AVERAGING
by Marco Martinolli 27 Feb '17

27 Feb '17

Dear all, I am new to FreeSurfer and I want to reproduce on 20 subjects the registration strategy as explained in Fischl paper: High-resolution inter-subject averaging and a coordinate system for the cortical surface, Fischl, B., Sereno, M.I., Tootell, R.B.H., and Dale, A.M., (1999). Human Brain Mapping, 8:272-284(1999) I found the page SurfaceRegAndTemplates - Free Surfer Wiki which is about the creation of a template for between-subject analysis but I am not sure if I have simply to use mris_make_template (in fact I do have the files spere.reg from the reconstruction step) or if I have to follow the 2-rounds procedure described in the page calling twice the make_average_subject. Which is the more correct one for my purpose? Right now, I tried with the second option but I encountered two different problems: 1. calling make_average_subject it works for 1 hour but at a certain moment it exit with errors saying that it does not find the file norm.mgz for the volume--> since I thought I was not interested in the volume registration, I tried to run the command with the --no-vol flag and in this way it works, but I am not sure this does not lead to any unpleasant consequence in the next stepsSo the command I used is: make_average_subject --out NEW_TEMPLATE_SURFACE --subjects s1 s2 s3 ... --no-vol 2. the mris_preprocess step (after having registered the subjects obtaining sphere.reg.newtemplate files) does not work as described in the page in the link. I have to use the input --meas thickness, the hemisphere --hemi lh/rh, the output file --out preproc-lh-thicknesss-mgh and especially the target flag --target NEW_TEMPLATE SURFACE. So the whole command line is: mris_preproc --surfreg sphere.reg.newtemplate --s s1 --s s2 --s s3... --meas thickness --hemi lh --out preproc-lh-thicknesss-mgh --target NEW_TEMPLATE_SURFACE Is it correct?I get the error that it does not find the target file. For instance when it takes subject 50002_recon it says that 50002_recon/surf/lh.NEW_TEMPLATE_SURFACE.sphere.reg does not exist.And it's true because it is in the subject folder NEW_TEMPLATE_SURFACE indeed and in the subjects folder I just have lh.sphere.reg.newtemplate. -------------------------------------------------------------------------------sysname Linuxhostname marco-Lenovo-Yoga-2-Promachine x86_64user marcosrcsubject = 50002_reconsrcval = /home/marco/Desktop/SUBJECTS_freeSurfer/50002_recon/surf/lh.thicknesssrctype = curvtrgsubject = NEW_TEMPLATE_SURFACEtrgval = ./tmp.mris_preproc.9144/50002_recon.1.mghtrgtype = srcsurfreg = NEW_TEMPLATE_SURFACE.sphere.regtrgsurfreg = sphere.reg.newtemplatesrchemi = lhtrghemi = lhframe = 0fwhm-in = 0fwhm-out = 0label-src = lh.cortex.labellabel-trg = lh.cortex.labelOKToRevFaceOrder = 1UseDualHemi = 0Reading source surface reg /home/marco/Desktop/SUBJECTS_freeSurfer/50002_recon/surf/lh.NEW_TEMPLATE_SURFACE.sphere.regNo such file or directorymri_surf2surf: could not read surface /home/marco/Desktop/SUBJECTS_freeSurfer/50002_recon/surf/lh.NEW_TEMPLATE_SURFACE.sphere.regNo such file or directory---------------------------------------------------------------------------------------------- I decided to skip the mris_preproc phase and end the procedure of the online documentation. So I run again make_average_subject with --no-vol and with the additional option --surf-regsphere.reg.newtemplate.The result is the FINAL_TEMPLATE folder.Now, observing the mgh file in the mri subfolder called "mni305.cor.mgz", the template appears to be quite blurred. Is it the template or something else? In order to have all the subjects aligned with this final template I have to run again the registrations as in the second step, right? I apologize for the huge amount of questions, but I am a lot confused and I spent a lot of time thinking what I should do.Please, any help you could give me would be highly appreciated, even more if you are clear in the explanation. Best regards,Marco

2 1

make average subject
by Limachia, Gaurang (NIH/NINDS) [F] 27 Feb '17

27 Feb '17

Hello Freesurfer Experts, I received this error while running the make_avg_subject script. I would greatly appreciate it if you can point me in the right direction. Also, I do not know how the aparc.annot file is supposed to be created. not aligning hemispheres before averaging. zeroing medial wall in aparc creating new parameterization... processing subject CON_201 (1 of 98) reading spherical surface /raids/hmcsraid3/silvina/PMD_FATCAT2015/CON_201/surf/lh.sphere.reg... could not read annot file /raids/hmcsraid3/silvina/PMD_FATCAT2015/CON_201/surf/../label/lh.aparc.annot No such file or directory mris_make_template: could not read annot file aparc No such file or directory ERROR: make_average_surface Thanks, Gaurang Shyam Limachia

2 1

Re: [Freesurfer] bbregister vs spmregister vs mri_coreg
by Antonin Skoch 27 Feb '17

27 Feb '17

Dear Doug, I was wondering, why do you prefer mni152reg to mri_coreg for gregistration to MNI? According to help, mri_coreg is also able to do 12 DOF registration. Antonin mri_coreg is the FS implementation of spm_coreg (spmregister) both of which use normalized mutual info. bbregister uses the BBR cost function and is preferred for all MRI. For registration to MNI space, we usually use mni152reg (a wrapper around fsl's flirt) On 2/25/17 8:49 AM, John Anderson wrote: Dear Freesurfer experts, I see that the tool "mri_coreg" has been implemented recently in Free Surfer 6 and I really wanted to know what are the differences between the registration tools "bbregister", " spmregister" and "mri_coreg"! Kindly: 1. Are these tools similar? if not what are the differences ? 2. Is there any preference of using a tool over the others for specific type of data. For example: A. If I want to register FA map to T1 image which tool is more robust? B. if I want to register FA map to MNI space which tool is more robust? I highly appreciate your input on this! John

3 3

Odd outputs from TRACULA
by Magnús Ingvi 27 Feb '17

27 Feb '17

Hello! I have gone through the TRACULA tutorial, and the tracts are pretty odd. I'm wondering if I did something incorrectly, but as far as I can tell, I followed the tutorial step by step. The individuals in question are healthy, no irregularities in the brains. I have attached images of the fmajor and fminor tracts as an example. I also attached the configuration file with descriptions of what I'm doing. Best M.Ingvi

1 0

Hippocampal Segmentation
by Jessica Sevick 27 Feb '17

27 Feb '17

Hello FreeSurfer Team , I was wondering if you could explain to me what the “Left-Hippocampus” and “Right-Hippocampus” labels in the subfield segmentation are? Are they the tail of the hippocampi? Or everything that is not assigned to one of the other seven labels? Additionally I am wondering how the hippocampal subfields are ordered in terms of their determination? Is it a particular order or is it randomized? Thanks, Jessica Sevick MRI Research Assistant Seaman Family MR Research Centre T: 403.944.8652 C: 587.576.6666

2 3

Re: [Freesurfer] Overestimation of volume in Hippocampus
by Antonin Skoch 27 Feb '17

27 Feb '17

Dear Michel, not sure if I understand. Do you mean that you cropped your T2 to FOV 256 mm ? I think that this should have no effect until you cropped parts of the brain. Maybe, as a part of quality-check you can check whether your T2 is correctly registered to T1 by inspection T1_to_<analysisID>.v10.QC.gif as stated in https://surfer.nmr.mgh.harvard.edu/fswiki/HippocampalSubfields Antonin Dear Antonin, Thank you for the assistance! I really appreciate it. My final goal is to determine the volume of the dentate gyrus however I first wanted to establish if the volume of the whole hippocampus was within expected boundaries before moving on to the dentate gyrus. Do you perhaps know if forcing the field of view to 256 would affect the segmentation because the T2 scans have a FOV of 320?Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Antonin Skoch <a...(a)ikem.cz> Verzonden: maandag 27 februari 2017 13:27 Aan: freesurfer(a)nmr.mgh.harvard.edu Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Dear Michel, by default in the recon-all the input images are resampled to 1x1x1mm3. To disable this and retain your resolution, you have to add -hires flag (and adapt the mris_inflate -n parameter). This could improve the segmentation results, however I would not expect that it is the reason of volume overestimation you are refering to. I am not expert at hippocampal anatomy, therefore I cannot comment on your results. But I am guessing, maybe the hippocampal atlas in FreeSurfer includes some portions which have not been included in the MCI studies you refer to? Maybe you can find answer in the Hippocampal subfields paper: http://www.nmr.mgh.harvard.edu/~iglesias/pdf/subfieldsNeuroimage2015preprin… Antonin Dear Antonin, I did use the T2 scan for the T2pial refinement during the recon-all however I actually already preformed the hippocampal subfields and when i looked at the txt file generated from that the results are quite similar. I used the multispectral segmentation. I didn't not use the -hires flag but what does this flag exactly do if i may ask? Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Antonin Skoch <a...(a)ikem.cz> Verzonden: maandag 27 februari 2017 12:31 Aan: freesurfer(a)nmr.mgh.harvard.edu Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Dear Michel, Given your resolution 0.8x0.8x0.8, did you run recon-all with -hires flag? What did you used T2 image for? As a -T2pial refinement? T2 image is not used for hippocampal segmentation in standard recon-all stream. But pay attention that with -hires you also have to alter default value for mris_inflate, as specified here http://freesurfer.net/fswiki/SubmillimeterRecon In my case -n was not enough, I would recommend -n 50 (more does not do any harm). In any case, I would recommend to try hippocampal subfields module. https://surfer.nmr.mgh.harvard.edu/fswiki/HippocampalSubfields This should be more precise than "standard" aseg segmentation and given you have also 3D T2, multispectral T1 and T2 segmentation can be used. Antonin Skoch Dear Shane, I have checked the aseg.mgz and in only one I would say the segmentation failed however the others look fine and I'm taking the measurements from the text file that is generated from the aseg.stats. Should I redo the recon-all with a different input? I have always given freesurfer T1 and a T2 scan to process both having a resolution of 0.8x0.8x0.8 mm. Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Shane S <shane.schofi...(a)yahoo.com> Verzonden: maandag 27 februari 2017 11:15 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Hi Michel Hu, Yes, that’s too large for MCI or even controls for that matter I think. Have you checked the aseg.mgz? I generally find that the hippocampus is well segmented, even in patients with a lot of atrophy. Are you taking the measurements with the asegstats2table? Sincerely, -- Shane S On 27 February 2017 at 10:12:09, Michel Hu (a...(a)live.nl<mailto:a...@live.nl>) wrote: Dear Shane, I apologize for not being clear. The reported value is of only one half of the hippocampus. I have only run 4 subjects of the 75 however they all have a reported value of between 6000 to 10000 mm^3 summed together for the hippocampus. Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Shane S <shane.schofi...(a)yahoo.com> Verzonden: maandag 27 februari 2017 11:00 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Hi Michel Hu, Is that a summed value of left and right hippocampus? That seems normal to me since averaged total hippocamppal volumes for MCI have been typically reported at around 2500-3500mm3. Thanks, -- Shane S On 27 February 2017 at 09:44:27, Michel Hu (a...(a)live.nl<mailto:a...@live.nl>) wrote: Dear Freesurfer experts, I have run several subjects through the recon-all pipeline including the hippocampal segmentation with freesurfer 6.0 on Ubuntu 16.04. However when looking at the text file with the volumes it seems that all the numbers are bigger than normal as all my subjects have MCI and freesurfer reports them having all over 5000 mm^3. I was wondering if the problem could lie somewhere in the recon-all process? Thank you for your time Sincerely, Michel Hu

2 1

Re: [Freesurfer] Overestimation of volume in Hippocampus
by Antonin Skoch 27 Feb '17

27 Feb '17

Dear Michel, by default in the recon-all the input images are resampled to 1x1x1mm3. To disable this and retain your resolution, you have to add -hires flag (and adapt the mris_inflate -n parameter). This could improve the segmentation results, however I would not expect that it is the reason of volume overestimation you are refering to. I am not expert at hippocampal anatomy, therefore I cannot comment on your results. But I am guessing, maybe the hippocampal atlas in FreeSurfer includes some portions which have not been included in the MCI studies you refer to? Maybe you can find answer in the Hippocampal subfields paper: http://www.nmr.mgh.harvard.edu/~iglesias/pdf/subfieldsNeuroimage2015preprin… Antonin Dear Antonin, I did use the T2 scan for the T2pial refinement during the recon-all however I actually already preformed the hippocampal subfields and when i looked at the txt file generated from that the results are quite similar. I used the multispectral segmentation. I didn't not use the -hires flag but what does this flag exactly do if i may ask?Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Antonin Skoch <a...(a)ikem.cz> Verzonden: maandag 27 februari 2017 12:31 Aan: freesurfer(a)nmr.mgh.harvard.edu Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Dear Michel, Given your resolution 0.8x0.8x0.8, did you run recon-all with -hires flag? What did you used T2 image for? As a -T2pial refinement? T2 image is not used for hippocampal segmentation in standard recon-all stream. But pay attention that with -hires you also have to alter default value for mris_inflate, as specified here http://freesurfer.net/fswiki/SubmillimeterRecon In my case -n was not enough, I would recommend -n 50 (more does not do any harm). In any case, I would recommend to try hippocampal subfields module. https://surfer.nmr.mgh.harvard.edu/fswiki/HippocampalSubfields This should be more precise than "standard" aseg segmentation and given you have also 3D T2, multispectral T1 and T2 segmentation can be used. Antonin Skoch Dear Shane, I have checked the aseg.mgz and in only one I would say the segmentation failed however the others look fine and I'm taking the measurements from the text file that is generated from the aseg.stats. Should I redo the recon-all with a different input? I have always given freesurfer T1 and a T2 scan to process both having a resolution of 0.8x0.8x0.8 mm. Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Shane S <shane.schofi...(a)yahoo.com> Verzonden: maandag 27 februari 2017 11:15 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Hi Michel Hu, Yes, that’s too large for MCI or even controls for that matter I think. Have you checked the aseg.mgz? I generally find that the hippocampus is well segmented, even in patients with a lot of atrophy. Are you taking the measurements with the asegstats2table? Sincerely, -- Shane S On 27 February 2017 at 10:12:09, Michel Hu (a...(a)live.nl<mailto:a...@live.nl>) wrote: Dear Shane, I apologize for not being clear. The reported value is of only one half of the hippocampus. I have only run 4 subjects of the 75 however they all have a reported value of between 6000 to 10000 mm^3 summed together for the hippocampus. Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Shane S <shane.schofi...(a)yahoo.com> Verzonden: maandag 27 februari 2017 11:00 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Hi Michel Hu, Is that a summed value of left and right hippocampus? That seems normal to me since averaged total hippocamppal volumes for MCI have been typically reported at around 2500-3500mm3. Thanks, -- Shane S On 27 February 2017 at 09:44:27, Michel Hu (a...(a)live.nl<mailto:a...@live.nl>) wrote: Dear Freesurfer experts, I have run several subjects through the recon-all pipeline including the hippocampal segmentation with freesurfer 6.0 on Ubuntu 16.04. However when looking at the text file with the volumes it seems that all the numbers are bigger than normal as all my subjects have MCI and freesurfer reports them having all over 5000 mm^3. I was wondering if the problem could lie somewhere in the recon-all process? Thank you for your time Sincerely, Michel Hu

2 1

Re: [Freesurfer] Overestimation of volume in Hippocampus
by Antonin Skoch 27 Feb '17

27 Feb '17

Dear Michel, Given your resolution 0.8x0.8x0.8, did you run recon-all with -hires flag? What did you used T2 image for? As a -T2pial refinement? T2 image is not used for hippocampal segmentation in standard recon-all stream. But pay attention that with -hires you also have to alter default value for mris_inflate, as specified here http://freesurfer.net/fswiki/SubmillimeterRecon In my case -n was not enough, I would recommend -n 50 (more does not do any harm). In any case, I would recommend to try hippocampal subfields module. https://surfer.nmr.mgh.harvard.edu/fswiki/HippocampalSubfields This should be more precise than "standard" aseg segmentation and given you have also 3D T2, multispectral T1 and T2 segmentation can be used. Antonin Skoch Dear Shane, I have checked the aseg.mgz and in only one I would say the segmentation failed however the others look fine and I'm taking the measurements from the text file that is generated from the aseg.stats. Should I redo the recon-all with a different input? I have always given freesurfer T1 and a T2 scan to process both having a resolution of 0.8x0.8x0.8 mm.Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Shane S <shane.schofi...(a)yahoo.com> Verzonden: maandag 27 februari 2017 11:15 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Hi Michel Hu, Yes, that’s too large for MCI or even controls for that matter I think. Have you checked the aseg.mgz? I generally find that the hippocampus is well segmented, even in patients with a lot of atrophy. Are you taking the measurements with the asegstats2table? Sincerely, -- Shane S On 27 February 2017 at 10:12:09, Michel Hu (a...(a)live.nl<mailto:a...@live.nl>) wrote: Dear Shane, I apologize for not being clear. The reported value is of only one half of the hippocampus. I have only run 4 subjects of the 75 however they all have a reported value of between 6000 to 10000 mm^3 summed together for the hippocampus. Sincerely, Michel Hu ________________________________ Van: freesurfer-boun...(a)nmr.mgh.harvard.edu <freesurfer-boun...(a)nmr.mgh.harvard.edu> namens Shane S <shane.schofi...(a)yahoo.com> Verzonden: maandag 27 februari 2017 11:00 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] Overestimation of volume in Hippocampus Hi Michel Hu, Is that a summed value of left and right hippocampus? That seems normal to me since averaged total hippocamppal volumes for MCI have been typically reported at around 2500-3500mm3. Thanks, -- Shane S On 27 February 2017 at 09:44:27, Michel Hu (a...(a)live.nl<mailto:a...@live.nl>) wrote: Dear Freesurfer experts, I have run several subjects through the recon-all pipeline including the hippocampal segmentation with freesurfer 6.0 on Ubuntu 16.04. However when looking at the text file with the volumes it seems that all the numbers are bigger than normal as all my subjects have MCI and freesurfer reports them having all over 5000 mm^3. I was wondering if the problem could lie somewhere in the recon-all process? Thank you for your time Sincerely, Michel Hu

2 1

Overestimation of volume in Hippocampus
by Michel Hu 27 Feb '17

27 Feb '17

Dear Freesurfer experts, I have run several subjects through the recon-all pipeline including the hippocampal segmentation with freesurfer 6.0 on Ubuntu 16.04. However when looking at the text file with the volumes it seems that all the numbers are bigger than normal as all my subjects have MCI and freesurfer reports them having all over 5000 mm^3. I was wondering if the problem could lie somewhere in the recon-all process? Thank you for your time Sincerely, Michel Hu

2 4

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Freesurfer