Hi Emma

As Martin said not having a control group (or placebo group in which subjects don't receive any treatment) is not a good study design.

In your case you will need to use LME because it looks like you are expecting some sort of non-linear change for the mean response over time (pre-treatment change vs post-treatment change). This could be modeled with a piece-wise lme model with a knot at the treatment point. Also you have several time points and you want to model the intra-subject correlations among those time points and variable variance across time points.

So, as your design may be more complex than the average I would recommend you to consult some local biostatistician and once you figure out the model then you can use the matlab lme tools for fitting the statistical model, computing F-statistics and making statistical inferences. The matlab code will be something like this (of course you need to change the function parameters):

1-Read your label eg.:

lhcortex = fs_read_label('freesurfer/subjects/fsaverage/label/lh.cortex.label');

2-Read the data file eg.:

[lhY, lhmri] = fs_read_Y('lh.thickness.mgh');

3-Fit a vertex-wise lme model with random effects.:

lhstats = lme_mass_fit_vw(X, [1 2], lhY, ni, lhcortex);

4-Perform vertex-wise inference eg.:

CM.C = [your contrast matrix];

F_lhstats = lme_mass_F(lhstats, CM);

5-Save results eg.:
fs_write_fstats(F_lhstats, lhmri,' sig.mgh', 'sig');

Best
-Jorge

De: Martin Reuter <mreuter@nmr.mgh.harvard.edu>
Para: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Enviado: Miércoles 24 de septiembre de 2014 8:55
Asunto: Re: [Freesurfer] longitudinal analysis and linear mixed effects models
Hi Emma,

yes, LME is Matlab, but the wiki page describe the use step-by-step.

LME makes sense in your case because some subjects have less time points than others.

The alternative (if only a very small minority of subjects differs from the rest) would be the 2-stage approach, where you first reduce the longitudinal data to a single estimate of change (the slope of a linear fit within-subject) and then compare these slopes across groups. However,
- this is not a good idea in general as it does not consider the fact that some subjects have less time points than others, which is especially problematic, if there is a bias across groups (if you did a group analysis)
- also, while QDEC could be used to compare this slope/atrophy measure across groups, it currently does not allow 'one sample group mean' to check if this slope is different from zero. You'd have to use mri_glmfit for that.

You can upgrade to 5.3 to use a more recent freeview and to run your post-processing (LME etc), but don't mix versions for the image processing part (recon-all).

And finally about the design in general. Not having a control group can be a big problem. For example just looking at atrophy rates in your treatment group will not tell you anything. You'll probably find some atrophy here and there, but is it different from no-treatment? Is it different from normal ageing? You would not be able to tell. But maybe your design is different, e.g. you could look for a correlation with drug dose etc.

I hope that helps.
Best, Martin
On 09/23/2014 05:55 PM, Emma Thompson wrote:
Hi FS,
I want to conduct a within-subjects longitudinal analysis, I thought I would be able to run a LME model in QDEC but this doesn't seem to be the case at all, is this correct? It seems I have to do everything using Matlab (big sigh).

https://surfer.nmr.mgh.harvard.edu/fswiki/LinearMixedEffectsModels

I have already preprocessed the data: 1) cross for all time points, 2) base, and then 3) long according to the awesome tutorial provided by FS.

I have only one group, a patient population that was scanned prior to (bseline) and at various time points (3 time points post) following treatment. Would you agree that the best analytical approach for me would be the LME model, especially since I have a few subjects with a couple of missing post-treatment time points?

Is there any way you would recommend another approach using QDEC?

Lastly, I have done everything in FS version 5.1, I'm considering upgrading to 5.3, since my freeview software seems to be out of date. It also seems that the Matlab tools I need to run LME are only available in FS version 5.2. At this point a little concerned since I'm hoping I didn't waste a bunch of time just realizing all this now, would upgrading to 5.3 negatively impact all the work I've done thus far using version 5.1?

Thanks for the help!!
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- 
Dr. Martin Reuter

Instructor in Neurology
  Harvard Medical School
Assistant in Neuroscience
  Dept. of Radiology, Massachusetts General Hospital
  Dept. of Neurology, Massachusetts General Hospital
Research Affiliate
  Computer Science and Artificial Intelligence Lab,
  Dept. of Electrical Engineering and Computer Science,
  Massachusetts Institute of Technology

A.A.Martinos Center for Biomedical Imaging
149 Thirteenth Street, Suite 2301
Charlestown, MA 02129

Phone: +1-617-724-5652
Email: 
   mreuter@nmr.mgh.harvard.edu
   reuter@mit.edu
Web  : http://reuter.mit.edu 
_______________________________________________
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.