[Freesurfer] FMRI group analysis in common surface space
Hi all, We want to perform a FMRI group analysis in common surface space. For that purpose, can we (1) run feat2surf on both cope and varcopes on first-level feat analysis (2) run higher-level feat on the output (using the transformed copes/varcopes as the inputs)? Thanks, Georg ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] cluster summary to MNI standard space
Hi, I'd like to extract %BOLD-change from clusters or highly activated voxels (first level fMRI data processed in FEAT, higher level done with mri_glmfit). Can you tell me what kind of space the cluster summary list created by mri_surfcluster refers to? Is it real Talairach space as TalX, TalY and TalZ implies it? Is there a simple method to transform those coordinates to the FSL standard space (MNI152)? That seems inevitable for me, since I use Peate (a featquery based GUI) to perform model related time series extraction. Or is it somehow possible to show the percentage signal change directly with freesurfer? Thanks in advance Georg ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] reg-feat2anat basics
Hi Doug! May be a very basic question for better understanding, but how much does reg-feat2anat care about the FSL preprocessing e.g. dewarping? And especially important for me, does it make a difference if you FNIRT or FLIRT on first level? Is it a good idea to process non-linearly registered subjects with reg-feat2anat? In other words: is anat2exf.register.dat actually affected by FEAT preprocessing output (warp field and transformation matrices), or not at all? Many thanks in advance Georg ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] mri_glmfit and dofs
Hi Doug! When performing surface-based group analysis with mri_glmfit, I have to specify the dofs of my feat runs. But when I correct for motion outliers on first level (by additional confound EVs identified by fsl_motion_outliers) there will be different dofs for every subject. Is a correct group analysis still possible? Which dof will be the right one then (the subject with the least dofs maybe)? Many thanks in advance Georg ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
AW: [Freesurfer] Registering FSL Feat output to the anatomical
Aha, then everything was fine, since it always showed up on the same side... Thanks Georg -Ursprüngliche Nachricht- Von: freesurfer-boun...@nmr.mgh.harvard.edu [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Douglas N Greve Gesendet: Donnerstag, 28. Mai 2009 18:07 An: Georg Homola Cc: 'freesurfer' Betreff: Re: [Freesurfer] Registering FSL Feat output to the anatomical Sorry, this one got lost. Those are "electronic fiducials" placed in the image near the col,row,slice=0. It is for helping you assess whether you have a left-right reversal. Georg Homola wrote: > Hi freesurfers, > > when manually checking the automatic registration of my test subject with > tkregister2, I see some weird artifact overlaid to the example_func. Nothing > of that can be seen in tkmedit or fslview. I attached some screenshots. > However, it doesn't seem to affect the results, but where does it come from? > > Thanks > Georg > > > > > > > > > > > > > > ___ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer -- Douglas N. Greve, Ph.D. MGH-NMR Center gr...@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422 In order to help us help you, please follow the steps in: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] mri_glmfit-sim bug?
Hi All (and especially Doug, again)! As one of the final steps I run mri_glmfit-sim for clusterwise correction (of my random effects analysis btw.). It always worked well and I also didn't change the command line recently, but since I updated Freesurfer to the latest version (4.3.1) it quits with: ERROR: you must supply --fwhm with --sim, even if it is 0 There is no flag like --fwhm for mri_glmfit-sim and when I try --fwhm-override it won't change anything. Am I missing something? Thanks Georg ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
AW: [Freesurfer] mri_glmfit-sim bug?
Yes, it's for a permutation test. Your fix seems to have solved the problem. No more errors so far. Thanks again. Georg -Ursprüngliche Nachricht- Von: freesurfer-boun...@nmr.mgh.harvard.edu [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Doug Greve Gesendet: Montag, 1. Juni 2009 01:46 An: Georg Homola Cc: freesurfer@nmr.mgh.harvard.edu Betreff: Re: [Freesurfer] mri_glmfit-sim bug? Is this for a permutation test? Sorry, this is my fault. I've put a new version here ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/mri_glmfit-sim that should fix the problem. Can you test it out? thanks doug On Mon, 1 Jun 2009, Georg Homola wrote: > Hi All (and especially Doug, again)! > > As one of the final steps I run mri_glmfit-sim for clusterwise correction > (of my random effects analysis btw.). It always worked well and I also > didn't change the command line recently, but since I updated Freesurfer to > the latest version (4.3.1) it quits with: > > ERROR: you must supply --fwhm with --sim, even if it is 0 > > There is no flag like --fwhm for mri_glmfit-sim and when I try > --fwhm-override it won't change anything. Am I missing something? > > Thanks > Georg > > > > ___ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > -- Douglas N. Greve, Ph.D. MGH-NMR Center gr...@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422 In order to help us help you, please follow the steps in: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] fsaverage across versions
Hi, my follow up question would be, how much has changed regarding fsaverage between freesurfer version 4.1.0 and 4.3.1? Is the added insula label the only change that has been made? Besides, when I load both annotations, the old and the new one, it seems to me almost every border of the parcellation has been shifted slightly. Is it how it's meant to be? I ask because I forgot to update the fsaverage in my separately stored Subjects_dir since version 4.1.0 and made a lot of analysis in the last months. To my mind it's good enough to rerun mri_surfcluster with the new fsaverage to get the updated annotations into my summary tables. Should that do the trick? Thanks again, Georg --- Iris, the registration targets were derived differently. With version 4, we automatically fill in all the ventricles. In verion 3, they were partially filled in manually. Version 3 creates a surface around the ventrical making it look like a sulcus, and this affects both the target and the registration to the target. So, on most of the surface, the registration will be very close, but it will deviate signifiacntly around the ventricles/medial wall. I think the fsfast in version 4 should work fine with the anatomicals (including fsaverage) of version 3. doug Steinmann, Iris wrote: Hi, we have several reconstructed brains, which were processed by freesurfer 3.0.5. We also kept using the fsaverage data from 3.0.5 in order to be consistent. We wish to do fMRI analysis with freesurfer/fs-fast 4.0.5 and found that talairach coordinates differ slightly when using the fsaverage from version 3.0.5 and the current fsaverage from 4.0.5. Yet, the both volumes appear pretty similar, the newer one maybe a little bit smoother. We would like to know what constitutes the exact difference between these two fsaverage datasets and whether it is possible to use the new version for analysis with the "old" fsaverage data without getting inconsistent results. Thanks a lot, Iris ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer -- Douglas N. Greve, Ph.D. MGH-NMR Center [EMAIL PROTECTED] Phone Number: 617-724-2358 Fax: 617-726-7422 In order to help us help you, please follow the steps in: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] fsaverage across versions
Can anybody confirm my assumption about fsaverage from last week? Is it ok to rerun mri_surfcluster with the new aparc.annot? Thanks a lot, Georg -- Von: freesurfer-boun...@nmr.mgh.harvard.edu [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Georg Homola Gesendet: Mittwoch, 1. Juli 2009 17:05 An: 'Freesurfer Mailing List' Betreff: Re: [Freesurfer] fsaverage across versions Hi, my follow up question would be, how much has changed regarding fsaverage between freesurfer version 4.1.0 and 4.3.1? Is the added insula label the only change that has been made? Besides, when I load both annotations, the old and the new one, it seems to me almost every border of the parcellation has been shifted slightly. Is it how it's meant to be? I ask because I forgot to update the fsaverage in my separately stored Subjects_dir since version 4.1.0 and made a lot of analysis in the last months. To my mind it's good enough to rerun mri_surfcluster with the new fsaverage to get the updated annotations into my summary tables. Should that do the trick? Thanks again, Georg --- Iris, the registration targets were derived differently. With version 4, we automatically fill in all the ventricles. In verion 3, they were partially filled in manually. Version 3 creates a surface around the ventrical making it look like a sulcus, and this affects both the target and the registration to the target. So, on most of the surface, the registration will be very close, but it will deviate signifiacntly around the ventricles/medial wall. I think the fsfast in version 4 should work fine with the anatomicals (including fsaverage) of version 3. doug Steinmann, Iris wrote: Hi, we have several reconstructed brains, which were processed by freesurfer 3.0.5. We also kept using the fsaverage data from 3.0.5 in order to be consistent. We wish to do fMRI analysis with freesurfer/fs-fast 4.0.5 and found that talairach coordinates differ slightly when using the fsaverage from version 3.0.5 and the current fsaverage from 4.0.5. Yet, the both volumes appear pretty similar, the newer one maybe a little bit smoother. We would like to know what constitutes the exact difference between these two fsaverage datasets and whether it is possible to use the new version for analysis with the "old" fsaverage data without getting inconsistent results. Thanks a lot, Iris ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer -- Douglas N. Greve, Ph.D. MGH-NMR Center [EMAIL PROTECTED] Phone Number: 617-724-2358 Fax: 617-726-7422 In order to help us help you, please follow the steps in: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] Three classes and GLM
Hi all, I'm having some problems getting a second level fixed effects analysis started on my subjects. There are three runs (dofsum = 884) in each case I'd like to compare. OSGM works fine so far, but when I apply a matrix like this: 1 0 0 0 1 0 0 0 1 To a contrast like that: 0.5 0.5 -1 mri_glmfit tells me: ERROR DOF = 0. I also tried it with an FSGD file, so my approach was the same as for the Fsgdf3G0V example (http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf3G0V). That makes no difference. When I divide the three runs into two classes, everything works fine again. But why the error with three classes? Any ideas? Thanks, Georg $ mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label $SUBJECTS_DIR/01/label/lh.cortex.label Reading source surface /data/freesurfer/01/surf/lh.white Number of vertices 149696 Number of faces299388 Total area 96879.085938 AvgVtxArea 0.647172 AvgVtxDist 0.880017 StdVtxDist 0.256396 $Id: mri_glmfit.c,v 1.138.2.10 2009/05/18 15:41:44 greve Exp $ cwd /data/freesurfer/higher/highest/fixed cmdline mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label /data/freesurfer/01/label/lh.cortex.label sysname Darwin FixVertexAreaFlag = 1 UseMaskWithSmoothing 1 OneSampleGroupMean 0 y/data/freesurfer/higher/highest/fixed/01/gender.lh.cope1.sm05.mgz logyflag 0 Xxmat.mat usedti 0 labelmask /data/freesurfer/01/label/lh.cortex.label maskinv 0 glmdir 01/results/gender.sm05.lh IllCondOK 0 DoFFx 1 FFxDOF 884 yFFxVar /data/freesurfer/higher/highest/fixed/01/gender.lh.varcope1.sm05.mgz Creating output directory 01/results/gender.sm05.lh Loading y from /data/freesurfer/higher/highest/fixed/01/gender.lh.cope1.sm05.mgz Loading yffxvar from /data/freesurfer/higher/highest/fixed/01/gender.lh.varcope1.sm05.mgz Saving design matrix to 01/results/gender.sm05.lh/Xg.dat Matrix condition is 1 /data/freesurfer/01/label/lh.cortex.label /data/freesurfer/01/label/lh.cortex.label Found 141245 points in label. Pruning voxels by thr: 0.00 Found 141245 voxels in mask Saving mask to 01/results/gender.sm05.lh/mask.mgh search space = 91409.8 DOF = 0 ERROR: DOF = 0 ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
AW: [Freesurfer] Three classes and GLM
Hi again, meanwhile I understand the GLM DOF is defined by the difference of inputs and classes (groups). Inputs - Groups = DOF So I need more inputs than groups, is that right? But that also means the g3v0.fsgd example (http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf3G0V) can't work with only three inputs, nor is it possible to compare two runs of only one subject (that would be g2v0.fsgd with only two inputs). Please correct me if I'm wrong, but are the following combinations not the same somehow? Design 1: 1 0 Contrast 1: 1 -1 1 0 0 1 Design 2: 1 0 0 Contrast 2: 0.5 0.5 -1 0 1 0 0 0 1 Since the DOF of the first design is 1, that could be a workaround for the second combination (with DOF = 0). But still, I can't imagine a possibility to compare only two inputs of the same or different subjects. Or am I totally off the track? Any comments would be appreciated! Thanks again, Georg -Ursprüngliche Nachricht- Von: freesurfer-boun...@nmr.mgh.harvard.edu [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Georg Homola Gesendet: Freitag, 17. Juli 2009 03:01 An: 'Freesurfer Mailing List' Betreff: [Freesurfer] Three classes and GLM Hi all, I'm having some problems getting a second level fixed effects analysis started on my subjects. There are three runs (dofsum = 884) in each case I'd like to compare. OSGM works fine so far, but when I apply a matrix like this: 1 0 0 0 1 0 0 0 1 To a contrast like that: 0.5 0.5 -1 mri_glmfit tells me: ERROR DOF = 0. I also tried it with an FSGD file, so my approach was the same as for the Fsgdf3G0V example (http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf3G0V). That makes no difference. When I divide the three runs into two classes, everything works fine again. But why the error with three classes? Any ideas? Thanks, Georg $ mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label $SUBJECTS_DIR/01/label/lh.cortex.label Reading source surface /data/freesurfer/01/surf/lh.white Number of vertices 149696 Number of faces299388 Total area 96879.085938 AvgVtxArea 0.647172 AvgVtxDist 0.880017 StdVtxDist 0.256396 $Id: mri_glmfit.c,v 1.138.2.10 2009/05/18 15:41:44 greve Exp $ cwd /data/freesurfer/higher/highest/fixed cmdline mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label /data/freesurfer/01/label/lh.cortex.label sysname Darwin FixVertexAreaFlag = 1 UseMaskWithSmoothing 1 OneSampleGroupMean 0 y/data/freesurfer/higher/highest/fixed/01/gender.lh.cope1.sm05.mgz logyflag 0 Xxmat.mat usedti 0 labelmask /data/freesurfer/01/label/lh.cortex.label maskinv 0 glmdir 01/results/gender.sm05.lh IllCondOK 0 DoFFx 1 FFxDOF 884 yFFxVar /data/freesurfer/higher/highest/fixed/01/gender.lh.varcope1.sm05.mgz Creating output directory 01/results/gender.sm05.lh Loading y from /data/freesurfer/higher/highest/fixed/01/gender.lh.cope1.sm05.mgz Loading yffxvar from /data/freesurfer/higher/highest/fixed/01/gender.lh.varcope1.sm05.mgz Saving design matrix to 01/results/gender.sm05.lh/Xg.dat Matrix condition is 1 /data/freesurfer/01/label/lh.cortex.label /data/freesurfer/01/label/lh.cortex.label Found 141245 points in label. Pruning voxels by thr: 0.00 Found 141245 voxels in mask Saving mask to 01/results/gender.sm05.lh/mask.mgh search space = 91409.8 DOF = 0 ERROR: DOF = 0 ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] Three classes and GLM / mri_glmfit dof bug
Hi Doug, so it should also be easily possible to compare only two runs of one subject as I initially wanted to do it. To compute the weighted combination would be a hell of a work for me, since I got almost 30 subjects with lots of copes and even more contrasts, so I would have to do that for more than a thousand variations manually (although partly scriptable). I'd rather wait for the bugfix, if that's not going to take too long. Or is there an older version of mri_glmfit without that bug I could use only for the fixed effects part of my analysis? A subsequent question for me is, how important is the exact number of dofs for mri_glmfit (or for the GLM in general) anyway? Since I got varying sums of dofs esp. because of motion outliers, I cumulated the dofs for every single case. The sums are in a range from 835 to 908. Now, could I just pass mri_glmfit (without that bug) the lowest number to stick with it, even though there are a bit more dofs in some cases, or would that change any results? Thank you very much for your help in advance! Best regards, Georg -Ursprüngliche Nachricht- Von: freesurfer-boun...@nmr.mgh.harvard.edu [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Douglas N Greve Gesendet: Freitag, 17. Juli 2009 20:04 An: Georg Homola Cc: 'Freesurfer Mailing List' Betreff: Re: [Freesurfer] Three classes and GLM This is a bug in mri_glmfit in that it does not recognize that the dof is really 884 and not 0. The easiest thing to do right now is to explicitly compute the weighted combination, then do an osgm. You can use fscalc.fsl to do this. doug Georg Homola wrote: > Hi all, > > I'm having some problems getting a second level fixed effects analysis > started on my subjects. There are three runs (dofsum = 884) in each case I'd > like to compare. > > OSGM works fine so far, but when I apply a matrix like this: > > 1 0 0 > 0 1 0 > 0 0 1 > > To a contrast like that: > > 0.5 0.5 -1 > > mri_glmfit tells me: ERROR DOF = 0. > > I also tried it with an FSGD file, so my approach was the same as for the > Fsgdf3G0V example (http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf3G0V). That > makes no difference. > > When I divide the three runs into two classes, everything works fine again. > But why the error with three classes? > > Any ideas? > > Thanks, > Georg > > > > $ mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar > 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir > 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label > $SUBJECTS_DIR/01/label/lh.cortex.label > Reading source surface /data/freesurfer/01/surf/lh.white > Number of vertices 149696 > Number of faces299388 > Total area 96879.085938 > AvgVtxArea 0.647172 > AvgVtxDist 0.880017 > StdVtxDist 0.256396 > > $Id: mri_glmfit.c,v 1.138.2.10 2009/05/18 15:41:44 greve Exp $ > cwd /data/freesurfer/higher/highest/fixed > cmdline mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar > 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir > 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label > /data/freesurfer/01/label/lh.cortex.label > sysname Darwin > FixVertexAreaFlag = 1 > UseMaskWithSmoothing 1 > OneSampleGroupMean 0 > y/data/freesurfer/higher/highest/fixed/01/gender.lh.cope1.sm05.mgz > logyflag 0 > Xxmat.mat > usedti 0 > labelmask /data/freesurfer/01/label/lh.cortex.label > maskinv 0 > glmdir 01/results/gender.sm05.lh > IllCondOK 0 > DoFFx 1 > FFxDOF 884 > yFFxVar /data/freesurfer/higher/highest/fixed/01/gender.lh.varcope1.sm05.mgz > Creating output directory 01/results/gender.sm05.lh > Loading y from > /data/freesurfer/higher/highest/fixed/01/gender.lh.cope1.sm05.mgz > Loading yffxvar from > /data/freesurfer/higher/highest/fixed/01/gender.lh.varcope1.sm05.mgz > Saving design matrix to 01/results/gender.sm05.lh/Xg.dat > Matrix condition is 1 > /data/freesurfer/01/label/lh.cortex.label > /data/freesurfer/01/label/lh.cortex.label > Found 141245 points in label. > Pruning voxels by thr: 0.00 > Found 141245 voxels in mask > Saving mask to 01/results/gender.sm05.lh/mask.mgh > search space = 91409.8 > DOF = 0 > ERROR: DOF = 0 > > > > ___ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] Three classes and GLM / mri_glmfit dof bug
Hi Doug, do you think it will be possible for you to fix that bug in a week or so? It's somewhat crucial for my analysis. Another thing I'd like to ask is: how feasible would it be to implement the --min operation into mri_concat? There is already a --max to compute the max of concatenated volumes, but I have to use fscalc.fsl to get the minimum of the inputs to calculate conjunctions (like Thomas Nichols recommended). So far so good. The problem I noticed is, since the surface has to be transformed into a volume and back, maybe due to some truncation effects the output surface is slightly different from the input (I dont mean the operation itself) and won't match to a mask anymore. In other words: the signal looks the same, but masking will fail. There are two options to get around this: doing the masking with fscalc.fsl or apply a dummy operation to the mask with fscalc.fsl to "shift" the mask the same way as the signal and then do the masking with mri_mask. Therefore I ask because the calculations with mri_concat would be much faster and it seems slightly more accurate, too. What do you think? Thanks again and sorry for the constant spamming... Georg -Ursprüngliche Nachricht- Von: Douglas N Greve [mailto:gr...@nmr.mgh.harvard.edu] Gesendet: Montag, 20. Juli 2009 21:10 An: Georg Homola Cc: 'Freesurfer Mailing List' Betreff: Re: [Freesurfer] Three classes and GLM / mri_glmfit dof bug It probably won't change it that much as it's just a t or F test. doug Georg Homola wrote: > Hi Doug, > > so it should also be easily possible to compare only two runs of one subject > as I initially wanted to do it. To compute the weighted combination would be > a hell of a work for me, since I got almost 30 subjects with lots of copes > and even more contrasts, so I would have to do that for more than a thousand > variations manually (although partly scriptable). > > I'd rather wait for the bugfix, if that's not going to take too long. Or is > there an older version of mri_glmfit without that bug I could use only for > the fixed effects part of my analysis? > > A subsequent question for me is, how important is the exact number of dofs > for mri_glmfit (or for the GLM in general) anyway? Since I got varying sums > of dofs esp. because of motion outliers, I cumulated the dofs for every > single case. The sums are in a range from 835 to 908. Now, could I just pass > mri_glmfit (without that bug) the lowest number to stick with it, even > though there are a bit more dofs in some cases, or would that change any > results? > > Thank you very much for your help in advance! > Best regards, > Georg > > > > -Ursprüngliche Nachricht- > Von: freesurfer-boun...@nmr.mgh.harvard.edu > [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Douglas N > Greve > Gesendet: Freitag, 17. Juli 2009 20:04 > An: Georg Homola > Cc: 'Freesurfer Mailing List' > Betreff: Re: [Freesurfer] Three classes and GLM > > This is a bug in mri_glmfit in that it does not recognize that the dof > is really 884 and not 0. The easiest thing to do right now is to > explicitly compute the weighted combination, then do an osgm. You can > use fscalc.fsl to do this. > > doug > > > > Georg Homola wrote: > >> Hi all, >> >> I'm having some problems getting a second level fixed effects analysis >> started on my subjects. There are three runs (dofsum = 884) in each case >> > I'd > >> like to compare. >> >> OSGM works fine so far, but when I apply a matrix like this: >> >> 1 0 0 >> 0 1 0 >> 0 0 1 >> >> To a contrast like that: >> >> 0.5 0.5 -1 >> >> mri_glmfit tells me: ERROR DOF = 0. >> >> I also tried it with an FSGD file, so my approach was the same as for the >> Fsgdf3G0V example (http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf3G0V). >> > That > >> makes no difference. >> >> When I divide the three runs into two classes, everything works fine >> > again. > >> But why the error with three classes? >> >> Any ideas? >> >> Thanks, >> Georg >> >> >> >> $ mri_glmfit --surf 01 lh --y 01/gender.lh.cope1.sm05.mgz --yffxvar >> 01/gender.lh.varcope1.sm05.mgz --ffxdof 884 --glmdir >> 01/results/gender.sm05.lh --X xmat.mat --C test.mat --label >> $SUBJECTS_DIR/01/label/lh.cortex.label >> Reading source surface /data/freesurfer/01/surf/lh.white >> Number of vertices 149696 >> Number of faces299388 >> Total area 96879.085938 >> AvgVtxArea 0.647172 >> AvgVtxDist 0.880017 >> StdVtxDist 0.25
[Freesurfer] mri_glmfit reference
Hi, following these refs I've found on the list and the mri_glmfit help, I assume the cluster-wise probability of the permutation simulation is based on cluster mass inference, is that right? Thanks, Georg Statistical Science 2004, Vol. 19, No. 4, 676-685 Institute of Mathematical Statistics, 2004 Permutation Methods: A Basis for Exact Inference Michael D. Ernst Human Brain Mapping 15:1-25(2001) Nonparametric Permutation Tests For Functional Neuroimaging: A Primer with Examples Thomas E. Nichols and Andrew P. Holmes Validating cluster size inference: random field and permutation methods Satoru Hayasaka and Thomas E. Nichols, NI, 20 (2003) 2343-2356. Hagler and Sereno, NI, 2006, vol 33 1093-1103. Smoothing and cluster thresholding for cortical surface-based group analysis of fMRI data ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
[Freesurfer] CUDA vs. ATI Stream / OpenCL
Hi Nick (and others), do you already have positive experience with ATI cards, too? We think about buying a new workstation with lots of GPU power so it would be great help to know what really pays off: GeForce GTX 480 or ATI Radeon HD 5870? Or should we go for Nvidia Tesla? What kind of tech specs are important here? Double or single precision floating point performance, number of cores, amount of memory... Thanks and cheers, Georg ___ 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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Re: [Freesurfer] CUDA vs. ATI Stream / OpenCL
Yes, I know, but will Freesurfer support ATI Stream and maybe OpenCL in the near future? OpenCL should be something that could run both on ATI and Nvidia GPUs. And then again, is it a good idea to invest money in Nvidia Tesla cards? Cheers, Georg Von: Pedro Paulo de Magalhães Oliveira Junior [mailto:p...@netfilter.com.br] Gesendet: Montag, 23. August 2010 00:08 An: Georg Homola Cc: freesurfer@nmr.mgh.harvard.edu Betreff: Re: [Freesurfer] CUDA vs. ATI Stream / OpenCL CUDA won't work on ATI. - Pedro Paulo de Magalhães Oliveira Junior Diretor de Operações Netfilter & SpeedComm Telecom -- <http://www.netfilter.com.br> www.netfilter.com.br -- For mobile: <http://www.netfilter.com.br/mobile> http://www.netfilter.com.br/mobile On Sun, Aug 22, 2010 at 19:04, Georg Homola wrote: Hi Nick (and others), do you already have positive experience with ATI cards, too? We think about buying a new workstation with lots of GPU power so it would be great help to know what really pays off: GeForce GTX 480 or ATI Radeon HD 5870? Or should we go for Nvidia Tesla? What kind of tech specs are important here? Double or single precision floating point performance, number of cores, amount of memory... Thanks and cheers, Georg ___ 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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. ___ 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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Re: [Freesurfer] Notes on CUDA Acceleration
Hi Richard and others, allow me one additional remark that may be crucial for those considering to invest in new cards. Although the Fermi class cards make use of the same architecture (Geforce GTX 480 and Tesla C2050 for example), for consumer products (GTX 400 series), double precision performance has been limited to a quarter of that of the "full" Fermi architecture (Tesla C20xx). Error checking and correcting memory (ECC) is also disabled on consumer cards. I don't really know how important double precision is for the CUDA enabled Freesurfer tools, but this could mean you have to buy four GTX cards to catch up with the performance of one Tesla card. Cheers, Georg -Ursprüngliche Nachricht- Von: freesurfer-boun...@nmr.mgh.harvard.edu [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] Im Auftrag von Richard G. Edgar Gesendet: Dienstag, 24. August 2010 15:38 An: freesurfer@nmr.mgh.harvard.edu Betreff: [Freesurfer] Notes on CUDA Acceleration Greetings, I've been asked to provide some extra information about GPU support in Freesurfer (being the one guilty of mri_em_register_cuda...). Firstly, there are no immediate plans for OpenCL support. It would be very nice to have - with ATI, NVIDIA _and_ x86 multicore backends. However, it's far less mature than CUDA. The good news is that the really 'hard' bit is restructuring the algorithms to fit well on a GPU. The syntax of CUDA and OpenCL is very similar (strange that), but OpenCL is more verbose. As for cards. for what is in the current release, any GeForce GTX-200 series or Tesla 10 series (i.e. C1060 and S1070) card should work (I don't know the Quadro model numbers - CUDA architecture 1.3 is the key feature). I think that everything should actually work on somewhat older cards, but the compile flags will have to be tweaked. So long as that threshold is reached, the only issue is the amount of RAM needed. Currently, I expect that any card with at least 1 GiB of RAM will have plenty, and the threshold for mri_em_register_cuda will be much lower than that. Going forward, I would strongly recommend purchasing 'Fermi' class cards. These are the GTX 400 series, and Tesla 20 series. The new architecture lifts some hardware limits on GPU kernels which are crippling for portions of mri_ca_register. With a more accelerated mri_ca_register, RAM limits may also come into play, until I can come up with a suitably cunning GPU implementation of the Gaussian Classifier Array (right now, I'm going to burn around 2 GiB on a single GCA, to make implementation simple). However, I have bigger fish to fry first. One final thing: Nick and I found last week that the accelerated mri_em_register_cuda doesn't seem to work prior to skull stripping. I'm going to work on this this week, but if you want to continue using the GPU accelerated binary, you'll have to turn off the FAST_TRANSLATION and FAST_TRANSFORM flags in mri_em_register.c, and recompile. This will increase the runtime to around 4 minutes on ernie, but will give results identical to the CPU code. I hope this is helpful, Richard ___ 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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. ___ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
