Hi Catherine,

with edits I think it is best to re-use the 5.0 cross data for the
longitudinal processing.

Also conceptually there is a difference between longitudinal processing
and longitudinal statistical analysis. 
The image processing will reduce variability of your measures and even
if you are interested in cross sectional differences between groups
(e.g. thickness differences at baseline), it would be a good idea to run
everything through the longitudinal stream. If there is longitudinal
image data, I think it will always be advantageous to process it with
the longitudinal stream as opposed to the cross sectional. 

Different scanners are usually not problematic if each subject stays on
their scanner and scanners are randomly assigned across groups. If
scanners get switched in a longitudinal study, many bad things can
happen:
- one group switches on average earlier than the other -> bias
- only a single or very few time points are on one side of the switch ->
cannot model the scanner effect well, as there is no reliable slope on
that side of the switch.
Anyway it may be a good idea to model the scanner as a covariate and
test for a scanner effect. We have seen effects even in software updates
on the scanner (e.g. growing brains) and certainly a hardware switch
will drastically change the images.

Linear mixed effects models are the right mechanism to analyze
longitudinal data.

Best, Martin



On Wed, 2013-01-09 at 19:22 +0000, Catherine Bois wrote:
> Thank you for your reply. I am currently trying to decide which method  
> would be the best to analyze the data we have...We have ca 500 scans  
> of individuals with differing number of time points for different  
> individuals and certain clinical measures we wish to relate to  
> particularly cortical thickness and different cortical regions.  
> However, the problem is that the scans were not all collected with the  
> same scanner; would longitudinal still prove to be advantageous over  
> cross-sectional analysis. As all the manual edits have been conducted  
> in version 5.0 it would be preferable to keep these and use version  
> 5.1 for -base -long (unless of course this could be done with better  
> consistency if using version 5.1?). I was also planning to analyze the  
> data with linear mixed effect models, does this sound reasonable?
> 
> Thanks very much for your help
> 
> 
> 
> 
> Quoting Martin Reuter <mreu...@nmr.mgh.harvard.edu> on Wed, 09 Jan  
> 2013 13:46:35 -0500:
> 
> > Hi Catherine,
> >
> > yes, the first step is independent (cross sectional) processing of all
> > time points (normal recon-all).
> >
> > We usually do not recommend to mix versions. But the 5.1 longitudinal
> > stream has been designed in a way that it can take the cross sectional
> > data from 5.0. So it is possible. Just make sure that if you add time
> > points and subjects you still use 5.0 for the cross sectional recons and
> > 5.1 for base and long to be consistent.
> >
> > Alternatively wait a little longer and re-run everything from scratch
> > with the new 5.2 version that will come out hopefully this month.
> >
> > Best, Martin
> >
> > On Wed, 2013-01-09 at 14:22 +0000, Catherine Bois wrote:
> >> Sorry, another question about longitudinal processing; just realized
> >> that my institution has been using version 5.0, so all the cross
> >> sectional processing has been done with this version, is it ok to go
> >> ahead and do base and long with version 5.1 once I have managed to get
> >> this installed?
> >>
> >> Thanks,
> >>
> >> ----- Forwarded message from c.b...@sms.ed.ac.uk -----
> >>     Date: Wed, 09 Jan 2013 13:35:12 +0000
> >>     From: Catherine Bois <c.b...@sms.ed.ac.uk>
> >>  Subject: longitudinal processing
> >>       To: freesurfer@nmr.mgh.harvard.edu
> >>
> >> Hi, just to clarify,
> >>
> >> When beginning longitudinal analysis, the first step is the same as
> >> cross-sectional processing, eg just the "normal" recon-all command,
> >> for each subject, and each individual time point? Then the norm.mgz
> >> files are used?
> >>
> >> Thanks
> >>
> >> --
> >> The University of Edinburgh is a charitable body, registered in
> >> Scotland, with registration number SC005336.
> >>
> >>
> >>
> >> ----- End forwarded message -----
> >>
> >>
> >
> >
> >
> >
> > 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.
> >
> 
> 
> 
> 


_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Reply via email to