Also, when I run the lme_mass_fit_EMinit, lme_mass_RgGrow, lme_mass_fit_Rgw,
I get a warning: Matrix is close to singular or badle scaled. Results may be inaccurate. RCOND=3e-21/ In parallel_function>make_general_channel/channel_general (line 914)
In remoteParallelFunction (line 38).
What could cause this warning? I am afraid my results will not be correct.
Swati
____________________________________________
Swati Rane Levendovszky, PhD
Interim Director,
Diagnostic Imaging Sciences Center
Assistant Professor,
Integrated Brain Imaging Center
Radiology
University of Washington
Box 357115 | AA035 HSB | Seattle, WA 98195
From: Swati Rane <srleven@uw.edu>
Date: Thursday, March 12, 2020 at 10:18 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] longitudinal FreeSurfer help
Hi Kersten,
Thanks for the reply. I think I am trying to understand the steps that follow.
Do I have to check whether my model with 2 random effects (intercept and time) is better than that with one (intercept)?
For the contrast I am interested in group wise longitudinal cortical thinning. And also, if the longitudinal thinning is different between groups.
Given that my X=[intercept time grp1 time*grp1 grp2 time*grp2 grp3 time*grp3 age@timepoint1 gender]
How do I set up CM? Do I need to run the lme_mass_F multiple time per contrast I provide?
Will my CM be
For Longitudinal thinning in grp 1: [1 0 0 1 0 0 0 0 0 0]
For Longitudinal thinning in grp 2: [1 0 0 0 0 1 0 0 0 0]
For Longitudinal thinning in grp 3: [1 0 0 0 0 0 0 1 0 0]
For Longitudinal thinning difference grp 2 – grp 1: [1 0 0 -1 0 1 0 0 0 0]?
Can you explain the CM set-up described in the tutorial?
Thank you.
Swati
From: <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of "Diers, Kersten /DZNE" <Kersten.Diers@dzne.de>
Reply-To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Date: Thursday, March 12, 2020 at 5:19 AM
To: "freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] longitudinal FreeSurfer help
External Email - Use Caution
your points 1-4 look fine, as far as I can see.
Regarding 5, the recommended spatiotemporal mass-univariate approach consists of calling three functions:
lme_mass_fit_EMinit, lme_mass_RgGrow, lme_mass_fit_Rgw
in that order. If that does not work, it is in my eyes possible to use the more simple mass-univariate approach using
Regarding 6, which specific questions do you have?
On Do, 2020-03-12 at 05:58 +0000, Swati Rane wrote:
External Email - Use Caution
Dear FreeSurfer experts,
I simply do not understand the LME setup for longitudinal analyses.
Here is what I have: 3 groups each with variable longitudinal data (in terms of umber of time points and interval between timepoints). I am interested in detecting if the rates of longitudinal atrophy in each
of the 3 groups and whether it is significantly different between the 3 groups.
- My Qdec file contains: fsid, fsid_base, time, age, group, gender
- I read in the relevant files with (output of surf2surf, lh.cortex, lh.sphere)
- I read the Qdec file removed the first column, obtained the subject IDs, keep numerical data and get a sorted ‘M’ matrix
- X=[intercept time grp1 time*grp1 grp2 time*grp2 grp3 time*grp3 age@timepoint1 gender]
- After that, should I use voxel wise mass-univariate model or spatiotemporal mass-univariate model
lme_mass_fit_vw or
lme_mass_fit_EMinit?
- After that I don’t understand the subsequent steps. Can you help me understand?
Any help is appreciated.
Thanks!
Swati
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer