External Email - Use Caution
Hello,
I'm analyzing a large dataset using an LME model to examine longitudinal changes in brain structure that vary as a function of group (patients vs controls). Because some participants were scanned several years apart in early childhood, the intra-subject registration failed in the longitudinal processing stream. For this reason, my data were not initially processed using the longitudinal processing stream within Freesurfer. However, per this thread - https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg55847.html - I still plan to use the LME model. As I understand it, this model assumes that the longitudinal processing stream was conducted. Therefore, I have the following questions:
1. would the "base" (aka within-subject template) for each subject be = time point 1? 2. would the "long run" for each subject be = each subject's individual time points? 3. assuming #1 and 2 are correct, would I need to set my directory structure to match that used in the longitudinal processing stream? For example, if I have 2 time points for subject 1234, and they are initially labeled "1234.time1" and "1234.time2", then in order to run the LME model, I would create a copy of "1234.time1" and rename the copy as "1234" to indicate to the software that this should be used as the base. And, I would also need to rename "1234.time1" and "1234.time2" as "1234.time1.long.1234" and "1234.time2.long.1234" to indicate that these folders should be used as the longitudinal time points.
Does this seem appropriate?
Thanks in advance for your help, Lara
Lara Foland-Ross, Ph.D. Senior Research Associate and Imaging Lab Manager Center for Interdisciplinary Brain Sciences Research Stanford University School of Medicine 401 Quarry Road, Room 1356 Stanford, CA 94305-5795
External Email - Use Caution
Hi there,
Reposting this message (below). Would appreciate any guidance you can provide.
Thanks, Lara
________________________________________ From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Lara Foland-Ross lfoland@stanford.edu Sent: Friday, February 7, 2020 12:46 PM To: freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] some basic questions regarding longitudinal analyses in Freesurfer
External Email - Use Caution
Hello,
I'm analyzing a large dataset using an LME model to examine longitudinal changes in brain structure that vary as a function of group (patients vs controls). Because some participants were scanned several years apart in early childhood, the intra-subject registration failed in the longitudinal processing stream. For this reason, my data were not initially processed using the longitudinal processing stream within Freesurfer. However, per this thread - https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg55847.html - I still plan to use the LME model. As I understand it, this model assumes that the longitudinal processing stream was conducted. Therefore, I have the following questions:
1. would the "base" (aka within-subject template) for each subject be = time point 1? 2. would the "long run" for each subject be = each subject's individual time points? 3. assuming #1 and 2 are correct, would I need to set my directory structure to match that used in the longitudinal processing stream? For example, if I have 2 time points for subject 1234, and they are initially labeled "1234.time1" and "1234.time2", then in order to run the LME model, I would create a copy of "1234.time1" and rename the copy as "1234" to indicate to the software that this should be used as the base. And, I would also need to rename "1234.time1" and "1234.time2" as "1234.time1.long.1234" and "1234.time2.long.1234" to indicate that these folders should be used as the longitudinal time points.
Does this seem appropriate?
Thanks in advance for your help, Lara
Lara Foland-Ross, Ph.D. Senior Research Associate and Imaging Lab Manager Center for Interdisciplinary Brain Sciences Research Stanford University School of Medicine 401 Quarry Road, Room 1356 Stanford, CA 94305-5795
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Lara,
no, you should not use time point 1 as base.
You need to distinguish between the image processing part and the statistics post processing analysis (LME). These are two separate things.
For the image processing in the longitudinal pipeline, we create the base (=subject template). The name base can be misleading and has nothing to do with "baseline". That is why we try to refer to it as "subject_template" in most documentation now. Anyway, one never analyzes results of the subject template, but of the final fine-tuned time points after the -long runs.
In your case, you can simply process the images independently (cross sectionally) and still perform the statistical LME analysis where you group results from the same subject.
For measures from the stats files (aseg, aparc...) you just obtain those from the cross dirs with asegstats2table. For surface results you simply map everything to fsaverage (mris_preproc) and then run the LME.
Best, Martin
On Thu, 2020-02-20 at 21:53 +0000, Lara Foland-Ross wrote:
External Email - Use CautionHi there,
Reposting this message (below). Would appreciate any guidance you can provide.
Thanks, Lara
From: freesurfer-bounces@nmr.mgh.harvard.edu < freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Lara Foland-Ross lfoland@stanford.edu Sent: Friday, February 7, 2020 12:46 PM To: freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] some basic questions regarding longitudinal analyses in Freesurfer
External Email - Use CautionHello,
I'm analyzing a large dataset using an LME model to examine longitudinal changes in brain structure that vary as a function of group (patients vs controls). Because some participants were scanned several years apart in early childhood, the intra-subject registration failed in the longitudinal processing stream. For this reason, my data were not initially processed using the longitudinal processing stream within Freesurfer. However, per this thread - https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg55847.html
- I still plan to use the LME model. As I understand it, this model
assumes that the longitudinal processing stream was conducted. Therefore, I have the following questions:
- would the "base" (aka within-subject template) for each subject be
= time point 1? 2. would the "long run" for each subject be = each subject's individual time points? 3. assuming #1 and 2 are correct, would I need to set my directory structure to match that used in the longitudinal processing stream? For example, if I have 2 time points for subject 1234, and they are initially labeled "1234.time1" and "1234.time2", then in order to run the LME model, I would create a copy of "1234.time1" and rename the copy as "1234" to indicate to the software that this should be used as the base. And, I would also need to rename "1234.time1" and "1234.time2" as "1234.time1.long.1234" and "1234.time2.long.1234" to indicate that these folders should be used as the longitudinal time points.
Does this seem appropriate?
Thanks in advance for your help, Lara
Lara Foland-Ross, Ph.D. Senior Research Associate and Imaging Lab Manager Center for Interdisciplinary Brain Sciences Research Stanford University School of Medicine 401 Quarry Road, Room 1356 Stanford, CA 94305-5795
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
External Email - Use Caution
Hi Martin,
Thanks. This makes a lot of sense. I'll merge my subjects' surface files that have been registered to fsaverage, then follow the rest of the protocol listed at https://surfer.nmr.mgh.harvard.edu/fswiki/LinearMixedEffectsModels.
One quick follow up question - when I use mris_preproc to merge thickness files, should I merge subjects' smoothed data (e.g., lh.thickness.fwhm15.fsaverage.mgh) and skip the smoothing step (run using mri_surf2surf)? Or should I merge unsmoothed data (e.g., lh.thickness.fwhm0.fsaverage.mgh) and then smooth once merged?
Thanks again, Lara
________________________________________ From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Martin Reuter mreuter@nmr.mgh.harvard.edu Sent: Friday, February 21, 2020 1:10 AM To: Freesurfer support list Subject: Re: [Freesurfer] some basic questions regarding longitudinal analyses in Freesurfer
Hi Lara,
no, you should not use time point 1 as base.
You need to distinguish between the image processing part and the statistics post processing analysis (LME). These are two separate things.
For the image processing in the longitudinal pipeline, we create the base (=subject template). The name base can be misleading and has nothing to do with "baseline". That is why we try to refer to it as "subject_template" in most documentation now. Anyway, one never analyzes results of the subject template, but of the final fine-tuned time points after the -long runs.
In your case, you can simply process the images independently (cross sectionally) and still perform the statistical LME analysis where you group results from the same subject.
For measures from the stats files (aseg, aparc...) you just obtain those from the cross dirs with asegstats2table. For surface results you simply map everything to fsaverage (mris_preproc) and then run the LME.
Best, Martin
On Thu, 2020-02-20 at 21:53 +0000, Lara Foland-Ross wrote:
External Email - Use CautionHi there,
Reposting this message (below). Would appreciate any guidance you can provide.
Thanks, Lara
From: freesurfer-bounces@nmr.mgh.harvard.edu < freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Lara Foland-Ross lfoland@stanford.edu Sent: Friday, February 7, 2020 12:46 PM To: freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] some basic questions regarding longitudinal analyses in Freesurfer
External Email - Use CautionHello,
I'm analyzing a large dataset using an LME model to examine longitudinal changes in brain structure that vary as a function of group (patients vs controls). Because some participants were scanned several years apart in early childhood, the intra-subject registration failed in the longitudinal processing stream. For this reason, my data were not initially processed using the longitudinal processing stream within Freesurfer. However, per this thread - https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg55847.html
- I still plan to use the LME model. As I understand it, this model
assumes that the longitudinal processing stream was conducted. Therefore, I have the following questions:
- would the "base" (aka within-subject template) for each subject be
= time point 1? 2. would the "long run" for each subject be = each subject's individual time points? 3. assuming #1 and 2 are correct, would I need to set my directory structure to match that used in the longitudinal processing stream? For example, if I have 2 time points for subject 1234, and they are initially labeled "1234.time1" and "1234.time2", then in order to run the LME model, I would create a copy of "1234.time1" and rename the copy as "1234" to indicate to the software that this should be used as the base. And, I would also need to rename "1234.time1" and "1234.time2" as "1234.time1.long.1234" and "1234.time2.long.1234" to indicate that these folders should be used as the longitudinal time points.
Does this seem appropriate?
Thanks in advance for your help, Lara
Lara Foland-Ross, Ph.D. Senior Research Associate and Imaging Lab Manager Center for Interdisciplinary Brain Sciences Research Stanford University School of Medicine 401 Quarry Road, Room 1356 Stanford, CA 94305-5795
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
Hi Lara,
if you are sure you will stick with 15fwhm, you can do either one. If you plan to also run analysis at different fwhm, you can first map and then smooth at different levels.
Either way, it should not make a big difference.
Best, Martin
On Fri, 2020-02-21 at 15:41 +0000, Lara Foland-Ross wrote:
External Email - Use CautionHi Martin,
Thanks. This makes a lot of sense. I'll merge my subjects' surface files that have been registered to fsaverage, then follow the rest of the protocol listed at https://surfer.nmr.mgh.harvard.edu/fswiki/LinearMixedEffectsModels.
One quick follow up question - when I use mris_preproc to merge thickness files, should I merge subjects' smoothed data (e.g., lh.thickness.fwhm15.fsaverage.mgh) and skip the smoothing step (run using mri_surf2surf)? Or should I merge unsmoothed data (e.g., lh.thickness.fwhm0.fsaverage.mgh) and then smooth once merged?
Thanks again, Lara
From: freesurfer-bounces@nmr.mgh.harvard.edu < freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Martin Reuter < mreuter@nmr.mgh.harvard.edu> Sent: Friday, February 21, 2020 1:10 AM To: Freesurfer support list Subject: Re: [Freesurfer] some basic questions regarding longitudinal analyses in Freesurfer
Hi Lara,
no, you should not use time point 1 as base.
You need to distinguish between the image processing part and the statistics post processing analysis (LME). These are two separate things.
For the image processing in the longitudinal pipeline, we create the base (=subject template). The name base can be misleading and has nothing to do with "baseline". That is why we try to refer to it as "subject_template" in most documentation now. Anyway, one never analyzes results of the subject template, but of the final fine-tuned time points after the -long runs.
In your case, you can simply process the images independently (cross sectionally) and still perform the statistical LME analysis where you group results from the same subject.
For measures from the stats files (aseg, aparc...) you just obtain those from the cross dirs with asegstats2table. For surface results you simply map everything to fsaverage (mris_preproc) and then run the LME.
Best, Martin
On Thu, 2020-02-20 at 21:53 +0000, Lara Foland-Ross wrote:
External Email - Use CautionHi there,
Reposting this message (below). Would appreciate any guidance you can provide.
Thanks, Lara
From: freesurfer-bounces@nmr.mgh.harvard.edu < freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Lara Foland- Ross lfoland@stanford.edu Sent: Friday, February 7, 2020 12:46 PM To: freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] some basic questions regarding longitudinal analyses in Freesurfer
External Email - Use CautionHello,
I'm analyzing a large dataset using an LME model to examine longitudinal changes in brain structure that vary as a function of group (patients vs controls). Because some participants were scanned several years apart in early childhood, the intra-subject registration failed in the longitudinal processing stream. For this reason, my data were not initially processed using the longitudinal processing stream within Freesurfer. However, per this thread - https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg55847.html
- I still plan to use the LME model. As I understand it, this
model assumes that the longitudinal processing stream was conducted. Therefore, I have the following questions:
- would the "base" (aka within-subject template) for each subject
be = time point 1? 2. would the "long run" for each subject be = each subject's individual time points? 3. assuming #1 and 2 are correct, would I need to set my directory structure to match that used in the longitudinal processing stream? For example, if I have 2 time points for subject 1234, and they are initially labeled "1234.time1" and "1234.time2", then in order to run the LME model, I would create a copy of "1234.time1" and rename the copy as "1234" to indicate to the software that this should be used as the base. And, I would also need to rename "1234.time1" and "1234.time2" as "1234.time1.long.1234" and "1234.time2.long.1234" to indicate that these folders should be used as the longitudinal time points.
Does this seem appropriate?
Thanks in advance for your help, Lara
Lara Foland-Ross, Ph.D. Senior Research Associate and Imaging Lab Manager Center for Interdisciplinary Brain Sciences Research Stanford University School of Medicine 401 Quarry Road, Room 1356 Stanford, CA 94305-5795
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 mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Lara Foland-Ross -
Martin Reuter