Freesurfer

freesurfer@nmr.mgh.harvard.edu

30132 discussions

Re: [Freesurfer] Freesurfer Digest, Vol 117, Issue 9
by Jonathan DuBois 07 Nov '13

07 Nov '13

Hi Doug, Sorry, the ${s} was typo, and the ${h} was for hemisphere, but I modified the commands below to make them more straight forward. With the first command I am running mris_preproc with multiple controls and one patient together and then running mri_surf2surf to do smoothing, and mri_glmfit with the fsgd and mtx files displayed below. In the second command I am running mris_preproc and mri_surf2surf on the control group and patient separately. Then I use mri_concat to get the mean and sd from the control group, which I compare to the patient file using fscalc.fsl to generate the zmap. I attached a screen shot from one subject to show how different the results are (fscalc method is the one with the large negative cluster in the prefrontal). The results of the fscalc.fsl method seem to be correct, but I am not sure why they differ so much. So my question is, if I use the fscalc.fsl method, is there a way to do the monte carlo based cluster correction? Alternatively, is there something wrong with the way I am running mri_glmfit that it is not running a z-test? Thanks again, Jonathan [image: Inline image 2][image: Inline image 1] ------------------------------ > > Message: 21 > Date: Tue, 05 Nov 2013 18:26:14 -0500 > From: Douglas N Greve <greve(a)nmr.mgh.harvard.edu> > Subject: Re: [Freesurfer] z-test differences from fscalc and > mri_glmfit > To: freesurfer(a)nmr.mgh.harvard.edu > Message-ID: <52797E96.704(a)nmr.mgh.harvard.edu> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > > When you say that you are getting very different results, what are you > comparing exactly? I can't tell what your mri_glmfit command is because > it is embedded in a script with lots of variables. Have mercy on me and > just give me a simple command line that I don't have to struggle with:) > doug > > > > > > On 11/05/2013 02:21 PM, Jonathan DuBois wrote: > > Hi All, > > > > In a previous post Doug suggested two methods for generating > > voxel-wise z-score maps, one using glmfit with one subject in one > > group and the second using fscalc.fsl. Link to post: > > > https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2010-June/014673.html > > > > > > I have tried both methods using PET data registered to the surface, > > but get very different results from fscalc and mri_glmfit. The results > > I expect are given by fscalc, but I would also like to run the Monte > > Carlo-based MCC, which requires a glmdir. Is there another way to run > > MC MCC on the z-map generated by fscal or alternatively, is the code > > that I am using for glmfit wrong? A simplified version of the code for > > both analyses is below. I am using FS 5.3 with the fsl binaries from > > the release notes on mac OS 10.8. > > > > > > Thanks for your help > > Jon > > > > ----------------------------------mri_glmfit: > > mris_preproc --target fsaverage --hemi lh --iv > > nc01_mrispace_pet.mgz nc01_mrispace_pet_2_T1_register.dat --iv > > nc02_mrispace_pet.mgz nc02_mrispace_pet_2_T1_register.dat --iv > > nc03_mrispace_pet.mgz nc03_mrispace_pet_2_T1_register.dat --iv > > pat01_mrispace_pet.mgz pat01_mrispace_pet_2_T1_register.dat --pf 0.5 > > --out lh.pat01_allcontrols_pet_ztest_pf05_fsaverage.mgh > > > > mri_surf2surf --hemi lh --s fsaverage --fwhm 10 > > --cortex --sval > > lh.pat01_allcontrols_pet_ztest_pf05_fsaverage.mgh --tval > > lh.pat01_allcontrols_pet_ztest_pf05_fsaverage_fwhm10.mgh; done > > > > mri_glmfit --y > > lh.pat01_allcontrols_pet_ztest_pf05_fsaverage_fwhm10.mgh --fsgd > > patient1_allcontrols_ztest.fsgd dods --C group2_cov0_diff.mtx --surf > > fsaverage lh --cortex --glmdir > > lh.pat01_allcontrols_pet_ztest_pf05_fsaverage_fwhm10.glmdir > > > > *patient1_allcontrols_ztest.fsgd*: > > GroupDescriptorFile 1 > > Title patient01_ztest > > Class Patient > > Class Control > > Input pat01 Patient > > Input nc01 Control > > Input nc02 Control > > Input nc03 Control > > > > *group2_cov0_diff.mtx*: > > -1 1 > > > > ----------------------------------fscalc.fsl: > > mris_preproc --target fsaverage --hemi lh --iv > > nc01_mrispace_pet.mgz nc01_mrispace_pet_2_T1_register.dat --iv > > nc02_mrispace_pet.mgz nc02_mrispace_pet_2_T1_register.dat --iv > > nc03_mrispace_pet.mgz nc03_mrispace_pet_2_T1_register.dat --pf 0.5 > > --out lh.allcontrols_pet_ztest_pf05_fsaverage.mgh > > > > mris_preproc --target fsaverage --hemi lh --iv > > pat01_mrispace_pet.mgz pat01_mrispace_pet_2_T1_register.dat --pf 0.5 > > --out lh.patient01_pet_ztest_pf05_fsaverage.mgh > > > > for stat in mean std; do mri_concat > > lh.allcontrols_pet_ztest_pf05_fsaverage.mgh --o > > lh.allcontrols_pet_ztest_pf05_fsaverage_${stat}.mgh --${stat}; done; > > done > > > > fscalc.fsl --surf fsaverage lh > > lh.patient01_pet_ztest_pf05_fsaverage.mgh -sub > > lh.allcontrols_pet_ztest_pf05_fsaverage_mean.mgh -div > > lh.allcontrols_pet_ztest_pf05_fsaverage_std.mgh > > lh.patient01_vs_allcontrols_pet_ztest_pf05_fsaverage_zmap.mgh; done > > > > > > _______________________________________________ > > Freesurfer mailing list > > Freesurfer(a)nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > -- > Douglas N. Greve, Ph.D. > MGH-NMR Center > greve(a)nmr.mgh.harvard.edu > Phone Number: 617-724-2358 > Fax: 617-726-7422 > > Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting > FileDrop: https://gate.nmr.mgh.harvard.edu/filedrop2 > www.nmr.mgh.harvard.edu/facility/filedrop/index.html > Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/ > > > > ------------------------------ > >

1 0

mris_fix_topology hangs (?)
by Martin Luessi 07 Nov '13

07 Nov '13

Hi, I'm running recon-all on a subject from the human connectome project, and for some reason "mris_fix_topology" seems to be hanging, or at least it has been using 100% of a CPU for more than 24h with no new output. Is this normal or is there a way to fix this problem? Best, Martin PS: I'm using FS 5.3 ... Correction of the Topology Finding true center and radius of Spherical Surface...done Surface centered at (0,0,0) with radius 100.0 in 28 iterations marking ambiguous vertices... 127679 ambiguous faces found in tessellation segmenting defects... 1 defects found, arbitrating ambiguous regions... analyzing neighboring defects... 1 defects to be corrected 0 vertices coincident reading input surface /autofs/cluster/fusion/mluessi/hcp_data/subjects/100307/surf/lh.qsphere.nofix... reading brain volume from brain... reading wm segmentation from wm... Computing Initial Surface Statistics -face loglikelihood: -10.9679 (-5.4839) -vertex loglikelihood: -10.8878 (-5.4439) -normal dot loglikelihood: -3.6700 (-3.6700) -quad curv loglikelihood: -8.7258 (-4.3629) Total Loglikelihood : -34.2514 CORRECTING DEFECT 0 (vertices=65102, convex hull=346)

3 4

MRISalloc: could not allocate vertices
by Ray, Siddharth 07 Nov '13

07 Nov '13

Hi Freesurfers, Got this error while running some of my subjects: XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Correcting Topology of defect 1 with euler number -257 (129 loops) computing statistics for defect 1: 33364 vertices location: [ (156,125,84) - average intensity = 92.126 ] -gray ( 86.57 , 6.29 ) -white ( 95.91 , 6.08 ) -gray ( 89.98 , 24.95 ) -white ( 98.58 , 24.03 ) -intensity (86.574921 [log = -2.404119 ]- 95.914825 [log = -2.517448 ]) -curv (k1=-0.162 (0.730) , r1 = 6.179 | k2=-0.060 (0.191), r2 = 16.806 ) -curv (k1=-0.124 (0.768) , r1 = 8.076 | k2=-0.032 (0.282), r2 = 31.306 ) MRISalloc(8654176, 520): could not allocate vertices Cannot allocate memory Do anyone know how to fix this? Thanks, -Siddharth

2 1

ltaReadFile error while running recon-all -long
by Jonathan Holt 07 Nov '13

07 Nov '13

FS experts, I’m having a bit of an issue with a single subject. Recon-all -long errors out after it tells me it cannot open the subjid_to_template.lta file. I’m very confused, because the file it’s looking for exists in the exact location it’s looking for but the recon-all.log suggests that there is no such file or directory. It proceeds to map control points and runs into a segmentation fault, after which it exits with errors. Any help would be appreciated.

1 1

Problems with autorecon2-cp
by Annemarie Brandhofe 07 Nov '13

07 Nov '13

Dear freesurfer community, we are conducting cortical-thickness analysis using freesurfer-5.3.0 and have been experiencing a problem with the automatic recognition of the control points. After setting the control points and running autorecon2-cp white matter segmentation is faulty in almost every slice. I attached screenshot of two slices. Can you recommend us a procedure to improve the automatic recognition of control points? Below, we provide the details of our setup. Feel free to ask for more information. Thank you very much for your help! Yours sincerely, Annemarie $ bugr FREESURFER_HOME: /apps/prod/i486/freesurfer-5.3.0 Build stamp: freesurfer-Linux-centos4-stable-pub-v5.3.0 Debian version: 5.0.6 Kernel info: Linux 2.6.27.39-1-ql-beowulf i686 $ which freesurfer -> /apps/prod/i486/freesurfer -5.3.0/bin/freesurfer Used command: recon-all –autorecon2-cp –subjid ET39_mdeft.

2 1

Literature values of cortical thickness
by Emma Biondetti 07 Nov '13

07 Nov '13

Dear Freesurfer experts, I need to compare the cortical thickness values, obtained from a data set of 32 healthy subjects, with some reference literature values of cortical thickness. I'd like to check the values of mean and standard deviation for each one of the 32 cortical parcellations automatically generated by Freesurfer. This is to decide if the data set can be used for the following steps of my work or not. On the mailing list, I read an old thread that talks about the existence of a set of literature reference values, obtained by Rauhl Desikan. Would it be possible to access those data? Thank you Emma Biondettu

2 1

Re: [Freesurfer] MRISread error
by Anita van Loenhoud 07 Nov '13

07 Nov '13

It finally worked! Thank you so much Louis. My SUBJECTS_DIR was pointing to a folder within my TUTORIAL_DATA directory (I thought it was supposed to be like that when I read the instructions), but now they point to the same folder and it works. Thanks again! Anita

1 0

A question as to subcortical volumetric measurement
by jh kim 07 Nov '13

07 Nov '13

Dear Doug and experts I've analyzed structural changes of the subcortical GM by using both VBM (DARTEL in SPM8) and subcortical volumemetric measurement using FreeSurfer 5.1. While there is no significant volumetric difference between the groups in VBM (corrected p<0.05), FS results showed that some subcortical GM had significant volume reductions in patients relative to controls (MANCOVA controlling for the effect of ICV, age, gender, followed by Bonferroni correction). One of the referees raised the question regarding this discrepancy between the VBM and FS findings. How can I cope with it? Thank you in advance for your help. Kim

3 5

unilateral hippocampus segmentation fail
by Andreas Berger 07 Nov '13

07 Nov '13

Hi FreeSurfers, i came across a few subjects in which FreeSurfer 5.3 failed to properly label most of the hippocampus in one hemisphere. Specifically, it is the left hippocampus in all of the 5 cases in which this occured. The subjects are young, healthy, and without any anatomical peculiarities. Segmentation was fine in these subjects with FreeSurfer 5.1. Screenshots of one of these subjects are attached. EDIT: that didn't seem to work, i uploaded them here: http://www.students.meduniwien.ac.at/~n0642496/files/freesurfer/freeview_hi… http://www.students.meduniwien.ac.at/~n0642496/files/freesurfer/freeview_hi… regards, Andreas

1 0

glm_fit --pvr
by celine＠nmr.mgh.harvard.edu 06 Nov '13

06 Nov '13

Dear Freesurfer team I am still trying to run a glm analysis between 2 surfaces but I am having weird results so I guess I missed something. So I sampled a DTI surface 2mm below the WM surface, and another intracortical surface (T2* actually) at 75% depth from the pial surface. The question I'd like to investigate is: is there a spatial relationship between DTI metrics below the WM surface and intracortical metrics here DTI in a patient population. So I run glm_fit with -pvr option. My design matrix has 2 classes (male and female) and 1 variable (age) that I entered as covariates. Here is the command: mri_glmfit --y lh.t2star_corr_disco_fsaverage__r0.9_depth0.75_fwhm3.mgh --pvr lh.L1_fsaverage_depth-2_fwhm5.mgh --fsgd ms_age_gender.fsgd --C lh_age_gender_pvr.mtx --surf fsaverage lh --glmdir t20.75_L1-2 --y input is the T2* surfaces for each subject concatenated -- pvr input is the DTI surface below the WM for each subject concatenated I wasn't so sure about the contrast, as I have 2 class, 1 variable + the pvr regressor I thought I would need 6 values (0 0 0 0 0 1) but I got an error about the contrast saying I should put 5 factors so my contrast is 0 0 0 0 1, to test the slope between T2* and DTI at each vertex, after taking into account gender and age. The output sig.mgh doesn't give me any significant result, even at the lowest possible threshold. So my first question would be: is it fine to correlate different surfaces that were not build at the same depth? And the 2nd question would be: is the contrast correct to test the correlation between the 2 surfaces at each vertex? Thanks a lot for your help Celine PS: here is the terminal output: gdfReadHeader: reading ms_age_gender.fsgd INFO: DeMeanFlag keyword not found, DeMeaning will NOT be done. Continuous Variable Means (all subjects) 0 Age 43.1176 9.02534 Class Means of each Continuous Variable 1 F 43.0435 2 M 43.2727 INFO: gd2mtx_method is dods Reading source surface /autofs/cluster/mscat/users/caterina/3T/recons_fsv5.3/fsaverage/surf/lh.white Number of vertices 163842 Number of faces 327680 Total area 65416.648438 AvgVtxArea 0.399267 AvgVtxDist 0.721953 StdVtxDist 0.195470 $Id: mri_glmfit.c,v 1.196.2.8 2012/11/01 18:51:41 greve Exp $ cwd /autofs/cluster/mscat/users/sindhuja/data/3T/results_2013-11-06_t2star_dti cmdline mri_glmfit --y lh.t2star_corr_disco_fsaverage__r0.9_depth0.75_fwhm3.mgh --pvr lh.L1_fsaverage_depth-2_fwhm5.mgh --fsgd ms_age_gender.fsgd --C lh_age_gender_pvr.mtx --surf fsaverage lh --glmdir t20.75_L1-2 sysname Linux hostname panarea machine x86_64 user celine FixVertexAreaFlag = 1 UseMaskWithSmoothing 1 OneSampleGroupMean 0 y /autofs/cluster/mscat/users/sindhuja/data/3T/results_2013-11-06_t2star_dti/lh.t2star_corr_disco_fsaverage__r0.9_depth0.75_fwhm3.mgh logyflag 0 usedti 0 FSGD ms_age_gender.fsgd labelmask /autofs/cluster/mscat/users/caterina/3T/recons_fsv5.3/fsaverage/label/lh.cortex.label maskinv 0 glmdir t20.75_L1-2 IllCondOK 0 ReScaleX 1 DoFFx 0 Creating output directory t20.75_L1-2 Loading y from /autofs/cluster/mscat/users/sindhuja/data/3T/results_2013-11-06_t2star_dti/lh.t2star_corr_disco_fsaverage__r0.9_depth0.75_fwhm3.mgh INFO: gd2mtx_method is dods Saving design matrix to t20.75_L1-2/Xg.dat Normalized matrix condition is 117.356 Matrix condition is 121295 Found 149955 points in label. Pruning voxels by thr: 0.000000 Found 2487 voxels in mask Saving mask to t20.75_L1-2/mask.mgh Reshaping mriglm->mask... search space = 1064.228206 DOF = 29 Starting fit and test Fit completed in 0.0014 minutes Computing spatial AR1 on surface Residual: ar1mn=0.602743, ar1std=0.415130, gstd=0.568767, fwhm=1.339345 Writing results lh_age_gender_pvr maxvox sig=4.08 F=20.9043 at index 63635 0 0 seed=1384618381 mri_glmfit done

1 0

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