Freesurfer

freesurfer@nmr.mgh.harvard.edu

1 participants
30133 discussions

Re: [Freesurfer] Longitudinal analysis within one group
by Martin Reuter 31 Jan '14

31 Jan '14

Given the different time distances, the best would be mixed effects models. The 2 stage model should also work as it incorporates time (but not different correlations according to time distance). Again I don't see what you are analyzing with only one group. There will always be thinning, and you will find it in some regions more than in others (depending on how much power you have). Best Martin Sent via my smartphone, please excuse brevity. -------- Original message -------- From: amirhossein manzouri <a.h.manzouri(a)gmail.com> Date:01/31/2014 7:27 AM (GMT-05:00) To: Martin Reuter <mreuter(a)nmr.mgh.harvard.edu> Cc: Douglas Greve <greve(a)nmr.mgh.harvard.edu>,free surfer <freesurfer(a)nmr.mgh.harvard.edu> Subject: Re: [Freesurfer] Longitudinal analysis within one group Dear Doug and Martin, Thanks a lot. I just have one group including 20 subjects. The scans have been done in a range of 0.9 to 1.5 years difference (tp2-tp1). Is ooh if I use the instructions from Paired Analysis :https://surfer.nmr.mgh.harvard.edu/fswiki/PairedAnalysis ? Regards, Best regards, Amirhossein Manzouri On Thu, Jan 30, 2014 at 10:31 PM, Martin Reuter <mreuter(a)nmr.mgh.harvard.edu> wrote: Hi, with one group you want to check if atrophy is significantly different from zero? That is probably the case for any group (e.g. aging), so it won't tell you anything really. Also, if you don't find atrophy in a region it doesn't mean it's not there (only your group size is too small to detect it). So the only real use I can think of, would be a test-retest study, where the assumption is that there is no change? Sadly Qdec cannot do 'one sample group mean', otherwise you could do that in qdec. If only you had a second 'control' group, then it could be done. You can use long_mris_slopes to compute rate or percent change maps (one for each subjects) and then you can use qdec to analyze those rate maps across groups. If you don't have a control group, you could still use long_mris_slopes (it can also map and stack the rate maps on fsaverage) and then simply run the one sample group mean in mri_glmfit. some info is also here: https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/LongitudinalTutorial_f… and https://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalTwoStageModel Best, Martin On 01/30/2014 09:18 AM, amirhossein manzouri wrote: Hi, Would you please advise if it is possible to do longitudinal statistical analysis within a group with two time points in Qdec. And if it is not possible in Qdec how I suppose to do it? Best regards, Amirhossein Manzouri _______________________________________________ Freesurfer mailing list Freesurfer(a)nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer -- Martin Reuter, Ph.D. Assistant in Neuroscience - Massachusetts General Hospital Instructor in Neurology - Harvard Medical School MGH / HMS / MIT A.A.Martinos Center for Biomedical Imaging 149 Thirteenth Street, Suite 2301 Charlestown, MA 02129 Phone: +1-617-724-5652 Email: mreuter(a)nmr.mgh.harvard.edu reuter(a)mit.edu Web : http://reuter.mit.edu 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.

1 0

Troubleshooting - uncertainty about adding control points
by Janosch Linkersdörfer 31 Jan '14

31 Jan '14

Dear Freesurfers, I am just getting started using Freesurfer. I ran recon-all on my subjects for the first time and now would like to troubleshoot the output. I watched the tutorial lectures and worked through the troubleshooting tutorial. Nevertheless, I am still uncertain which things I will have to correct (especially as editing too much and/or not to the same degree between subjects might do more harm than good). Some errors are relatively obvious like the exclusion of white matter voxels from the white matter surface (see *1* on the screen shot of my own data) I am far more uncertain about adding control points. For example, I do not know whether I should add those at the location denoted with *2* in my own data. If so, there would be a lot of other, similar locations where I would add control points in my data. But this would also be true for the image used in the "Adding control points" tutorial. In the second screen shot, which displays this image, I have marked some locations where the white matter surface does not follow the one I would draw if I had to do it manually. Is one supposed to correct all those imprecisions by adding control points?! What are the criteria for adding vs. not adding control points? Thank you very much, Janosch

2 3

Re: [Freesurfer] longitudinal processing, template and flags
by Martin Reuter 31 Jan '14

31 Jan '14

Hi Daniel, I'd also include 24 month into both the processing and analysis. It is not really that we are trying to improve a registration here. We are trying to improve segmentation and surface placement of each time point. I'd aways use all data, except there are good reasons to drop (eg strong motion artifacts). Best Martin Sent via my smartphone, please excuse brevity. -------- Original message -------- From: Daniel Ferreira <danifer(a)ull.es> Date:01/31/2014 3:41 AM (GMT-05:00) To: freesurfer@nmr.mgh.harvard.edu,mreuter@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] longitudinal processing, template and flags Thanks Martin for your reply, I am agree with you on point 1. What would you say in the case that we are interested in rates of decline from baseline to 12 months, but besides 6 months follow-up we also have 24 months?. Here I see more conceptual problems. In an hypothetical context of severe atrophy from 12 to 24 months, adding the 24 mo follow-up perhaps difficult registration more than improves it. What do you think? Thanks a lot Best Daniel Hi Daniel, 1. yes. But why not include 6 month in your study if you have those scans??? You will get much more reliable slope estimates with 3 time points compared to 2, so including that time point will not only help for the image processing part, but also in your statistics (you gain power). 2. I am not familiar with the mprage flag and think it mainly affects the normalization. If you are sure you need it, then also add it to the -base and -long runs, just to be safe. best, Martin On 01/29/2014 09:08 AM, Daniel Ferreira wrote: Dear experts, I have two questions regarding the longitudinal stream: - I am including Baseline and 12 months follow-up in my study. However, I also have 6 months follow-up. I wonder if it is conceptually right to include these 6 months scans in template creation with the idea of increasing registration. - I am applying the -mprage flag in the crossectional first step. Should I use it in the -base and -longs steps too? Thanks very much Daniel Ferreira

2 1

Longitudinal analysis within one group
by amirhossein manzouri 31 Jan '14

31 Jan '14

Hi, Would you please advise if it is possible to do longitudinal statistical analysis within a group with two time points in Qdec. And if it is not possible in Qdec how I suppose to do it? Best regards, Amirhossein Manzouri

3 3

Re: [Freesurfer] longitudinal processing, template and flags
by Daniel Ferreira 31 Jan '14

31 Jan '14

Thanks Martin for your reply, I am agree with you on point 1. What would you say in the case that we are interested in rates of decline from baseline to 12 months, but besides 6 months follow-up we also have 24 months?. Here I see more conceptual problems. In an hypothetical context of severe atrophy from 12 to 24 months, adding the 24 mo follow-up perhaps difficult registration more than improves it. What do you think? Thanks a lot Best Daniel Hi Daniel, 1. yes. But why not include 6 month in your study if you have those scans??? You will get much more reliable slope estimates with 3 time points compared to 2, so including that time point will not only help for the image processing part, but also in your statistics (you gain power). 2. I am not familiar with the mprage flag and think it mainly affects the normalization. If you are sure you need it, then also add it to the -base and -long runs, just to be safe. best, Martin On 01/29/2014 09:08 AM, Daniel Ferreira wrote: Dear experts, I have two questions regarding the longitudinal stream: - I am including Baseline and 12 months follow-up in my study. However, I also have 6 months follow-up. I wonder if it is conceptually right to include these 6 months scans in template creation with the idea of increasing registration. - I am applying the -mprage flag in the crossectional first step. Should I use it in the -base and -longs steps too? Thanks very much Daniel Ferreira

1 0

-hemi flag and control points
by silver47＠gmx.com 30 Jan '14

30 Jan '14

Hi, If control points are only added to one hemisphere, is it ok to use the -hemi flag together with -autorecon2-cp -autorecon3 for recon-all? Thanks, P

2 1

tksurfer overlay color range control bug report
by Joseph Dien 30 Jan '14

30 Jan '14

Hi, I’m using FreeSurfer 5.3.0 on OS X 10.9.1. I seem to have run into a GUI bug. When I go to Configure Overlay Display and change the Min setting the display range changes appropriately on the figure once I press the Apply button. When I change the max setting (e.g., from 5 to 50) the Histogram changes appropriately when I press Apply (the green line moves to 50) but nothing changes on the figure. Furthermore, if the color scale is displayed, the scale ends are unchanged (still 5). If I press Close and then open it up again the Max is no longer changed (still 5) and the Histogram now displays the unchanged value (green line at 5). If, however, I change the Max by right-clicking then not only does the histogram change (without needing to press Apply) but also the figure once I press Apply. Cheers! Joe Cut-and-paste the following info into your FreeSurfer problem report: --------------------------------------------------------------------- FREESURFER_HOME: /Applications/freesurfer Build stamp: freesurfer-Darwin-lion-stable-pub-v5.3.0 Kernel info: Darwin 13.0.0 x86_64 --------------------------------------------------------------------- Please include the following additional information in your report: 1) tksurfer overlay color range control bug report 2) tksurfer fsaverage lh inflated -aparc -overlay AP1-fusiform-voxel-ST-FacSF1lh.nii.gz 3) none 4) not relevant See also: http://surfer.nmr.mgh.harvard.edu/fswiki/BugReporting which tksurfer /Applications/freesurfer/bin/tksurfer -------------------------------------------------------------------------------- Joseph Dien, Senior Research Scientist Maryland Neuroimaging Center University of Maryland E-mail: jdien07(a)mac.com Phone: 202-297-8117 http://joedien.com

1 0

Re: [Freesurfer] Longitudinal RepeatedMeasuresAnova mri_glmfit "--surf subject hemi"
by Martin Reuter 30 Jan '14

30 Jan '14

Hi Kaiming, 1. Depends on what you are interested in (atrophy per month or yearly rate , select 1 or 2). To include age I'd make a column with age at baseline or mean age of the subject's time points, not time varying. I have no clue what you are looking at, but the significance maps should not change . But you really don't have a analysis here anyway with only one subject! So who knows what's going on. 2. Usually people look at thickness or volume. Selecting area or area_mm grabs those columns from the stats files. I have no idea what the difference is . Best Martin Sent via my smartphone, please excuse brevity. -------- Original message -------- From: Kaiming Yin <yinkaiming6987(a)gmail.com> Date:01/30/2014 6:21 PM (GMT-05:00) To: Martin Reuter <mreuter(a)nmr.mgh.harvard.edu> Cc: FreeSurfer Mailing List <freesurfer(a)nmr.mgh.harvard.edu> Subject: Re: [Freesurfer] Longitudinal RepeatedMeasuresAnova mri_glmfit "--surf subject hemi" Hi Martin, Thanks for the help of glm, I also try the QDEC table for the 3 times points of 1 same subject k. k was scanned at Aug-2012, May-2013, and Jul-2013. A little confusion here is should I record the time as which one of these: (1) --time years (2) --time months (3) --time ages 0 0 25.36 0.77 9 26.13 0.92 11 26.28 I used these three timing ways and find that the freeview overlay shreshold differ a lot. both the (1) --time years and (3) --time ages can use threshold range as 2, 5 which i think means p<0.01 and p>0.00001? but the (2) --time months fits the threshold only as 0.1, 0.5 (p<0.8 and p>0.3?). The shape of the histogram are all the same in all these three and the area covered are also the same. In the tutorial I see the example use the first timing way, i.e. --time years, which also use the threshold as 2,5. Another question is when I call the long_stats_slopes command: long_stats_slopes --qdec long.qdec.table.dat --stats aseg.stats --meas Area_mm2 --do-avg --do-rate --do-pc1 --do-spc --do-stack --time ages --sd ../.. long_stats_slopes --qdec long.qdec.table.dat --stats lh.aparc.stats --meas area --do-avg --do-rate --do-pc1 --do-spc --do-stack --time ages --sd ../.. the flag --meas uses two different but very similar measures for area, and not sure what is the difference there. Thanks a lot, Kaiming On Thu, Jan 30, 2014 at 6:46 PM, Martin Reuter <mreuter(a)nmr.mgh.harvard.edu> wrote: Hi Kaiming, thickness is not a surface, it is a function defined on a surface. Don't add it as a surface name to the command line. Best, Martin On 01/30/2014 11:12 AM, Kaiming Yin wrote: Dear Martin, Make sense now. thanks for this. Yes I only have one subject with 3 different time points, what about the longitudinal two stage model (QDEC)? will this work? and one more question about the glm is when doing the thickness comparison, calling the mri_glmfit command, should I mention: mri_glmfit ... --surf fsaverage lh thickness? or just left it as ... --surf fsaverage lh? I saw the default <surfacename> is 'white'. But when I add this 'thickness' parameter it seems can not work either and reports like: " gdfReadHeader: reading rmanova.fsgd INFO: DeMeanFlag keyword not found, DeMeaning will NOT be done. Continuous Variable Means (all subjects) 0 TP1-vs-TP2 0 0.816497 1 TP1-vs-TP3 0 0.816497 Class Means of each Continuous Variable 1 k 0.0000 0.0000 INFO: gd2mtx_method is doss Reading source surface /Freesurfer//k/surf/lh.thickness nquads=5767172, nvertices=569 ERROR: MRISread: file '/Freesurfer//k/surf/lh.thickness' has many more faces than vertices! Probably trying to use a scalar data file as a surface! Invalid argument " Thanks for your patience, Kaiming On Thu, Jan 30, 2014 at 3:46 PM, Martin Reuter <mreuter(a)nmr.mgh.harvard.edu> wrote: Dear Kaiming, how many subjects do you have? only 1? You need to have more data to run a glm (your degrees of freedom is 0 , see error message). Your other question: you can select any target for running your analysis. But usually people do not select one of the subjects from the study, because that may introduce a processing bias towards that subject (maybe the thickness changes of that subject will drive the results or bias it). That is why fsaverage is recommended. Best, Martin On 01/30/2014 06:13 AM, Kaiming Yin wrote: Dear Martin, Yes I think so, and I run the mri_glmfit with fsaverage as the subject like: mri_glmfit --glmdir lh.thickness.rmanova --y lh.thickness.sm05.mgh --fsgd rmanova.fsgd doss --C tp1-vs-tp2.mtx --C tp1-vs-tp3.mtx --C tp2-vs-tp3.mtx --C tp-effect.mtx --C mean.mtx --surf fsaverage lh it reports: FixVertexAreaFlag = 1 UseMaskWithSmoothing 1 OneSampleGroupMean 0 y /Freesurfer/k/lh.thickness.sm05.mgh logyflag 0 usedti 0 FSGD rmanova.fsgd labelmask /Freesurfer//fsaverage/label/lh.cortex.label maskinv 0 glmdir lh.thickness.rmanova IllCondOK 0 ReScaleX 1 DoFFx 0 Creating output directory lh.thickness.rmanova Loading y from /Freesurfer/k/lh.thickness.sm05.mgh INFO: gd2mtx_method is doss Saving design matrix to lh.thickness.rmanova/Xg.dat Normalized matrix condition is 1.73205 Matrix condition is 1.73205 Found 149955 points in label. Pruning voxels by thr: 0.000000 Found 149955 voxels in mask Saving mask to lh.thickness.rmanova/mask.mgh Reshaping mriglm->mask... search space = 74612.965197 DOF = 0 ERROR: DOF = 0 It seems still something wrong was ignored. Also if I call the preproc or surf2surf commands with the base subject, i.e. k, instead of fsaverage, will it work? or even better? mris_preproc --target k --hemi lh --meas thickness --out lh.thickness.mgh --fsgd rmanova.fsgd mri_surf2surf --s k --hemi lh --fwhm 5 --sval lh.thickness.mgh --tval lh.thickness.sm05.mgh Thanks, Kaiming On Thu, Jan 30, 2014 at 3:18 AM, Martin Reuter <mreuter(a)nmr.mgh.harvard.edu> wrote: Your data was probably mapped to fsaverage when you call the preproc or surf2surf commands? If so you should specify fsaverage as the subject nane. Best Martin Sent via my smartphone, please excuse brevity. -------- Original message -------- From: Kaiming Yin Date:01/29/2014 9:07 PM (GMT-05:00) To: FreeSurfer Mailing List Subject: [Freesurfer] Longitudinal RepeatedMeasuresAnova mri_glmfit "--surf subject hemi" Dear FreeSurfer guys, I am using longitudinal processing stream and with the RepeatedMeasuresAnova. I follow the instruction here from http://surfer.nmr.mgh.harvard.edu/fswiki/RepeatedMeasuresAnova my fsgd file is: GroupDescriptorFile 1 Title k_longitudinal Class k Variables TP1-vs-TP2 TP1-vs-TP3 Input k_1 k 1 1 Input k_2 k -1 0 Input k_3 k 0 -1 contrasts is: tp1-vs-tp2.mtx tp1-vs-tp3.mtx tp2-vs-tp3.mtx tp-effect.mtx mean.mtx 0 1 0 0 0 1 0 -1 1 0 1 0 1 0 0 0 0 1 I run the mris_preproc and mri_surf2surf commands as the same for left hemi thickness, but when I run the mri_glmfit command it reports "ERROR: you must use '--surface subject hemi' with surface data". if i add the flag as: mri_glmfit ...(as the same before)... --surface k lh or --surface k lh thickness or --surface k_1.long.k lh or --surface k_1.long.k lh thickness or --surface k_1 lh or --surface k_1 lh thickness it still goes wrong and even not make an output folder. Any problems there? thanks, Kaiming 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. -- Martin Reuter, Ph.D. Assistant in Neuroscience - Massachusetts General Hospital Instructor in Neurology - Harvard Medical School MGH / HMS / MIT A.A.Martinos Center for Biomedical Imaging 149 Thirteenth Street, Suite 2301 Charlestown, MA 02129 Phone: +1-617-724-5652 Email: mreuter(a)nmr.mgh.harvard.edu reuter(a)mit.edu Web : http://reuter.mit.edu -- Martin Reuter, Ph.D. Assistant in Neuroscience - Massachusetts General Hospital Instructor in Neurology - Harvard Medical School MGH / HMS / MIT A.A.Martinos Center for Biomedical Imaging 149 Thirteenth Street, Suite 2301 Charlestown, MA 02129 Phone: +1-617-724-5652 Email: mreuter(a)nmr.mgh.harvard.edu reuter(a)mit.edu Web : http://reuter.mit.edu

1 0

transferring annotation from one subject to another
by Marie Schaer 30 Jan '14

30 Jan '14

Hi all, I have an annotation for fsaverage that I would like to convert into a subject's space. Up to now, the only way I found to do that was to convert my annotation to all its labels, and then rebuilt the annotation in the subject's space from the labels, using the following 3 steps mri_annotation2label --annotation myAparc --hemi ?h --subject fsaverage --outdir fsaverage/label/ mri_label2label --srcsubject fsaverage --srclabel fsaverage/label/?h.label1.label --trgsubject SUBJ --regmethod surface --hemi ?h --trglabel SUBJ/label/?h.label1.label (which needs to be repeated for each label) mris_label2annot --ctab myColorLUT.txt --l ?h.label1.label [ … up to --?h.labelX.label] --s SUBJ --a myAparc --hemi ?h That is however not really efficient for annotations including increasing numbers of labels. Is there any other way to convert more efficiently the annotation, something like an annotation2annotation command? Thanks a lot in advance for your answer, Marie

1 1

Re: [Freesurfer] Longitudinal RepeatedMeasuresAnova mri_glmfit "--surf subject hemi"
by Martin Reuter 30 Jan '14

30 Jan '14

Your data was probably mapped to fsaverage when you call the preproc or surf2surf commands? If so you should specify fsaverage as the subject nane. Best Martin Sent via my smartphone, please excuse brevity. -------- Original message -------- From: Kaiming Yin <yinkaiming6987(a)gmail.com> Date:01/29/2014 9:07 PM (GMT-05:00) To: FreeSurfer Mailing List <freesurfer(a)nmr.mgh.harvard.edu> Subject: [Freesurfer] Longitudinal RepeatedMeasuresAnova mri_glmfit "--surf subject hemi" Dear FreeSurfer guys, I am using longitudinal processing stream and with the RepeatedMeasuresAnova. I follow the instruction here from http://surfer.nmr.mgh.harvard.edu/fswiki/RepeatedMeasuresAnova my fsgd file is: GroupDescriptorFile 1 Title k_longitudinal Class k Variables TP1-vs-TP2 TP1-vs-TP3 Input k_1 k 1 1 Input k_2 k -1 0 Input k_3 k 0 -1 contrasts is: tp1-vs-tp2.mtx tp1-vs-tp3.mtx tp2-vs-tp3.mtx tp-effect.mtx mean.mtx 0 1 0 0 0 1 0 -1 1 0 1 0 1 0 0 0 0 1 I run the mris_preproc and mri_surf2surf commands as the same for left hemi thickness, but when I run the mri_glmfit command it reports "ERROR: you must use '--surface subject hemi' with surface data". if i add the flag as: mri_glmfit ...(as the same before)... --surface k lh or --surface k lh thickness or --surface k_1.long.k lh or --surface k_1.long.k lh thickness or --surface k_1 lh or --surface k_1 lh thickness it still goes wrong and even not make an output folder. Any problems there? thanks, Kaiming

2 5

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