[Freesurfer] mri_glmfit application in the two-stage longitudinal approach
Hi Martin, Now I move onto the mri_glmfit step (still within the two stage method). The on-line tutorial for doing this is not very clear. The output of the first stage is stored in the subject-template (base), already in fsaverage space. Simply replace the thickness file with your long.thickness-pc1fit (etc) and run your analysis. For the normal group analysis, one mgh file (the stacked image for all subject for each hemisphere) is created under the SUBJECT folder after the mri_prepro, and the mri_surf2surf steps. However, I assume that after the long_mris_slopes command, the mgh file (lh.long.thickness-spc.fwhm5.fsaverage.mgh, etc.) saved within each subject's base/surf folder is just for that subject. Right? My question is: what is the right way to process these subject specific mgh files within mri_glmfit? How to design the FSGD file for this particular glmfit? I appreciate your clarification on this issue, and I do believe that it would be very helpful if you could add your clarification onto the online tutorial for the two stage method ( http://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalTwoStageModel) Many thanks Keith On Wed, May 8, 2013 at 4:28 PM, Martin Reuter mreu...@nmr.mgh.harvard.eduwrote: Hi Keith, your longitudinal qdec file looks OK. Something must have gone wrong before (in the processing, e.g. maybe the base was reprocessed and the longs were not updated or things were renamed?) I would create a new directory with only this subject, then start from scratch. create both cross (kp5 and kp5b), the base (kp5_base) and both longs with the correct naming . Then check if the long_mris_slopes command still fails. Best, Martin On 05/01/2013 12:11 PM, Qi Wu wrote: Dear Freesurfer experts, I have a dataset of 2 time points. I ran into problems when I tried to use the two-stage approach to pull out statistics such as the rate of change in Freesurfer 5.1. I could not even get the first stage working. The error messages I got were “ERROR: MRISar1: Surf/Src dimension mismatch” “ERROR -11 : mri_glmfit did not work?” I did further trouble-shooting for this particular message nv1 = 140942, nv1 = 143525, and I found out that 143525 was in the base folder (kp5_base), and 140942 was for the two longitudinally registered images (kp5.long.kp5_base, and kp5b.long.kp5_base). The command line I used was : “long_mris_slopes --qdec ./qdec/AS_long_qdec.dat --meas thickness --hemi lh --do-avg --do-rate --do-pc1 --do-spc --do-stack --do-label --time weeks –nosmooth”. The first three lines in the qdec file: “fsid fsid-base weeks age sex FSS avgPain kp5 kp5_base 0 48 0 4.1 7 kp5b kp5_base 20 48 0 8.6 6” This error was similar to what was in a previous post ( https://mail.nmr.mgh.harvard.edu/pipermail/freesurfer/2012-June/024576.htmlby joost janssen Jun 29, 2012). I tried to fix any similar problems, but still wasn’t able to get it working. What else could I have been doing wrong? The on screen error output is provided as follows Many thanks Keith Postdoctoral fellow Toronto Western Hospital Toronto, Ontario Canada *** Parsing the qdec table: ./qdec/AS_long_qdec.dat Working in SUBJECTS_DIR: /data/keith/freesurferdata/subjects_longitudinal_5.1_old Subject-Template: kp5_base INFO: 2 TPs in kp5_base , mean age: 10.0 === SUBJECT kp5_base Intersecting Within-Subject Cortex Label cp /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5.long.kp5_base/label/lh.cortex.label /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/label/lh.long.cortex.label mris_label_calc intersect /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5b.long.kp5_base/label/lh.cortex.label /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/label/lh.long.cortex.label /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/label/lh.long.cortex.label LabelWrite: saving to /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/label/lh.long.cortex.label === SUBJECT kp5_base Stackinbg Within-Subject Maps mri_concat /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5.long.kp5_base/surf/lh.thickness /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5b.long.kp5_base/surf/lh.thickness --o /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/surf/lh.long.thickness-stack.mgh ninputs = 2 Checking inputs nframestot = 2 Allocing output Done allocing nframes = 2 Writing to /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/surf/lh.long.thickness-stack.mgh === SUBJECT kp5_base Running Within-Subject GLM Writing ./tmp-kp5_base_lh_thickness_6PR2rH/X-long.mat ...
Re: [Freesurfer] mri_glmfit application in the two-stage longitudinal approach
Hi Keith, you are right, the lh.long.thickness-spc.fwhm5.fsaverage.mgh in each base folder is only for that subject. You process this the same way as you'd usually do thickness in a cross sectional analysis. Except you can skip the mris_prepoc step (as everyting is mapped to fsaverage and presmoothed already, when using the qcache flag to long_mris_slopes). You have several options - simply use mri_concat to stack those files to create the stack - or (probably easier) tell long_mris_slopes to do the job for you with the --stack-rate or --stack-spc etc. options Then you have a stack on fsaverage with the right smoothing level(s). the FSGD file is also as usual, make sure the order of your subjects agrees with the order in the long.qdec table used for the stacking. Added that to the wiki , thanks. Best, Martin On 05/21/2013 02:36 PM, Qi Wu wrote: Hi Martin, Now I move onto the mri_glmfit step (still within the two stage method). The on-line tutorial for doing this is not very clear. The output of the first stage is stored in the subject-template (base), already in fsaverage space. Simply replace the thickness file with your long.thickness-pc1fit (etc) and run your analysis. For the normal group analysis, one mgh file (the stacked image for all subject for each hemisphere) is created under the SUBJECT folder after the mri_prepro, and the mri_surf2surf steps. However, I assume that after the long_mris_slopes command, the mgh file (lh.long.thickness-spc.fwhm5.fsaverage.mgh, etc.) saved within each subject's base/surf folder is just for that subject. Right? My question is: what is the right way to process these subject specific mgh files within mri_glmfit? How to design the FSGD file for this particular glmfit? I appreciate your clarification on this issue, and I do believe that it would be very helpful if you could add your clarification onto the online tutorial for the two stage method (http://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalTwoStageModel) Many thanks Keith On Wed, May 8, 2013 at 4:28 PM, Martin Reuter mreu...@nmr.mgh.harvard.edu mailto:mreu...@nmr.mgh.harvard.edu wrote: Hi Keith, your longitudinal qdec file looks OK. Something must have gone wrong before (in the processing, e.g. maybe the base was reprocessed and the longs were not updated or things were renamed?) I would create a new directory with only this subject, then start from scratch. create both cross (kp5 and kp5b), the base (kp5_base) and both longs with the correct naming . Then check if the long_mris_slopes command still fails. Best, Martin On 05/01/2013 12:11 PM, Qi Wu wrote: Dear Freesurfer experts, I have a dataset of 2 time points. I ran into problems when I tried to use the two-stage approach to pull out statistics such as the rate of change in Freesurfer 5.1. I could not even get the first stage working. The error messages I got were “ERROR: MRISar1: Surf/Src dimension mismatch” “ERROR -11 : mri_glmfit did not work?” I did further trouble-shooting for this particular message nv1 = 140942, nv1 = 143525, and I found out that 143525 was in the base folder (kp5_base), and 140942 was for the two longitudinally registered images (kp5.long.kp5_base, and kp5b.long.kp5_base). The command line I used was : “long_mris_slopes --qdec ./qdec/AS_long_qdec.dat --meas thickness --hemi lh --do-avg --do-rate --do-pc1 --do-spc --do-stack --do-label --time weeks –nosmooth”. The first three lines in the qdec file: “fsid fsid-base weeks age sex FSS avgPain kp5 kp5_base 0 48 0 4.1 7 kp5b kp5_base 20 48 0 8.6 6” This error was similar to what was in a previous post (https://mail.nmr.mgh.harvard.edu/pipermail/freesurfer/2012-June/024576.html by joost janssen Jun 29, 2012). I tried to fix any similar problems, but still wasn’t able to get it working. What else could I have been doing wrong? The on screen error output is provided as follows Many thanks Keith Postdoctoral fellow Toronto Western Hospital Toronto, Ontario Canada *** Parsing the qdec table: ./qdec/AS_long_qdec.dat Working in SUBJECTS_DIR: /data/keith/freesurferdata/subjects_longitudinal_5.1_old Subject-Template: kp5_base INFO: 2 TPs in kp5_base , mean age: 10.0 === SUBJECT kp5_base Intersecting Within-Subject Cortex Label cp /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5.long.kp5_base/label/lh.cortex.label /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5_base/label/lh.long.cortex.label mris_label_calc intersect /data/keith/freesurferdata/subjects_longitudinal_5.1_old/kp5b.long.kp5_base/label/lh.cortex.label
