I think we must be careful about including noisy large scale measurements as global nuisance covariates in the General Linear Model (GLM). The GLM assumes that the independent variables are measured almost without error (eg. see http://en.wikipedia.org/wiki/Errors-in-variables_models).%C2%A0 For instance, mean cortical thickness or mean global activity in fMRI (as the mean of many values) should not be as noisy as Intracranial Volume estimates (or the estimated volume of any neuroanatomical structure).
-Jorge
De: Anderson Winkler andersonwinkler@hotmail.com Para: freesurfer@nmr.mgh.harvard.edu Enviado: Lunes 26 de marzo de 2012 13:56 Asunto: Re: [Freesurfer] Cortical Normalization Questions
Hi Jeff and all,
For normalization (i.e., divide the measurement under study by some global measurement), I would not argue favourably, as this procedure can bias the results in the opposite direction if a global effect is present.
Instead, include it as a covariate is not as harmful. My suggestion is, when there is no clear approach
about using or not a global measurement
as a nuisance, the relationships between the measurement under study, the independent variable, and the putative nuisance should be calculated, and models with and without the nuisance should be analysed and presented. The discussion should consider both analyses together, and enough information should be presented so that the final interpretation is left to the reader.
Specifically for area, I suggest analysing and presenting two models: (1) without any global measurement and (2) with global area as nuisance.
If brain volume (whichever way it is measured) is to be considered a potential nuisance for the disorder you are analysing, it can be included in the model #2 above, even given that they are not orthogonal to each other, and are related to global area. Non-orthogonality between the nuisance variables is not a problem as it is when effects of interest are involved.
Hope this helps!
All the best,
Anderson
On 23/03/12 11:29, Michael Harms wrote:
Our reply to that is here http://bjp.rcpsych.org/content/196/5/414.2.long
which reminded me of other papers that have also used a global thickness measure to covary for mean cortical thickness and thereby "address whether any regional thickness differences were in excess of global cortical thickness differences between groups" -- see references [1,4] in our Reply.
cheers, -MH
Hi Michael and others,
maybe it's this one:
http://bjp.rcpsych.org/content/196/5/414.1.long
best,
-joost
On Fri, Mar 23, 2012 at 2:15 AM, Michael Harms mharms@conte.wustl.eduwrote:
Hi Jeff, I personally like the idea of using average thickness as a covariate to control for a reduction in "whole brain" thickness, and have used that approach in a paper. If the Abstract that you mentioned indicated that this is flawed, I'd be curious to know what the reason was...
cheers, -MH
On Thu, 2012-03-22 at 21:00 -0400, Bruce Fischl wrote:
Hi Jeff
yes, I think this is still our recommendation for thickness, although perhaps David Salat can verify. As far as surface area,
you might get
Anderson Winkler to send you a preprint of his newly accepted paper on surface area comparisons and how to do them properly. I would have
said
normalize by the 2/3 root of ICV (maybe David can comment on this as
well)
cheers Bruce
On Thu, 22 Mar 2012, Jeff Sadino wrote:
> Hello, > For cortical thickness normalizations, Bruce said not to normalize
based on a HBM
> abstract > (
http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg06646.html). Is
> this still the
consensus?
> > For cortical volume, it is pretty standard to normalize to eTIV. > > For cortical surface area (jacobian), I couldn't find any
information on the wiki.
> Does anyone have any recommendations? > > Thank you, > Jeff > > _______________________________________________ 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.
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 Jorge,
The global measurement enters in the model as a nuisance, not as an effect of interest. Even if the estimator is inconsistent, it is still useful as such. About which is noisier, it turns out that in general, the bigger the structure, the less measurement noise is present. From the structural measurements we use, the global ones are certainly those with higher SNR, including brain volumes, average cortical thickness and total area -- even if, e.g. brain volume is based on a single measurement rather than on some sort of average.
All the best,
Anderson
On 27/03/12 01:50, jorge luis wrote:
I think we must be careful about including noisy large scale measurements as global nuisance covariates in the General Linear Model (GLM). The GLM assumes that the independent variables are measured almost without error (eg. see http://en.wikipedia.org/wiki/Errors-in-variables_models). For instance, mean cortical thickness or mean global activity in fMRI (as the mean of many values) should not be as noisy as Intracranial Volume estimates (or the estimated volume of any neuroanatomical structure).
-Jorge
------------------------------------------------------------------------ *De:* Anderson Winkler <andersonwinkler@hotmail.com> *Para:* freesurfer@nmr.mgh.harvard.edu *Enviado:* Lunes 26 de marzo de 2012 13:56 *Asunto:* Re: [Freesurfer] Cortical Normalization Questions Hi Jeff and all, For normalization (i.e., divide the measurement under study by some global measurement), I would not argue favourably, as this procedure can bias the results in the opposite direction if a global effect is present. Instead, include it as a covariate is not as harmful. My suggestion is, when there is no clear approach about using or not a global measurement as a nuisance, the relationships between the measurement under study, the independent variable, and the putative nuisance should be calculated, and models with and without the nuisance should be analysed and presented. The discussion should consider both analyses together, and enough information should be presented so that the final interpretation is left to the reader. Specifically for area, I suggest analysing and presenting two models: (1) without any global measurement and (2) with global area as nuisance. If brain volume (whichever way it is measured) is to be considered a potential nuisance for the disorder you are analysing, it can be included in the model #2 above, even given that they are not orthogonal to each other, and are related to global area. Non-orthogonality between the nuisance variables is not a problem as it is when effects of interest are involved. Hope this helps! All the best, Anderson On 23/03/12 11:29, Michael Harms wrote: > Our reply to that is here > http://bjp.rcpsych.org/content/196/5/414.2.long > > which reminded me of other papers that have also used a global thickness > measure to covary for mean cortical thickness and thereby "address whether > any regional thickness differences were in excess of global cortical > thickness differences between groups" -- see references [1,4] in our > Reply. > > cheers, > -MH > >> Hi Michael and others, >> >> maybe it's this one: >> >> http://bjp.rcpsych.org/content/196/5/414.1.long >> >> best, >> -joost >> >> >> On Fri, Mar 23, 2012 at 2:15 AM, Michael Harms >> <mharms@conte.wustl.edu <mailto:mharms@conte.wustl.edu>>wrote: >> >>> Hi Jeff, >>> I personally like the idea of using average thickness as a covariate to >>> control for a reduction in "whole brain" thickness, and have used that >>> approach in a paper. If the Abstract that you mentioned indicated that >>> this is flawed, I'd be curious to know what the reason was... >>> >>> cheers, >>> -MH >>> >>> On Thu, 2012-03-22 at 21:00 -0400, Bruce Fischl wrote: >>>> Hi Jeff >>>> >>>> yes, I think this is still our recommendation for thickness, although >>>> perhaps David Salat can verify. As far as surface area, you might get >>>> Anderson Winkler to send you a preprint of his newly accepted paper on >>>> surface area comparisons and how to do them properly. I would have >>> said >>>> normalize by the 2/3 root of ICV (maybe David can comment on this as >>> well) >>>> cheers >>>> Bruce >>>> >>>> >>>> On Thu, 22 Mar 2012, Jeff Sadino wrote: >>>> >>>>> Hello, >>>>> For cortical thickness normalizations, Bruce said not to normalize >>> based on a HBM >>>>> abstract >>>>> ( >>> http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg06646.html). >>> Is >>>>> this still the consensus? >>>>> >>>>> For cortical volume, it is pretty standard to normalize to eTIV. >>>>> >>>>> For cortical surface area (jacobian), I couldn't find any >>> information >>> on the wiki. >>>>> Does anyone have any recommendations? >>>>> >>>>> Thank you, >>>>> Jeff >>>>> >>>>> >>>> _______________________________________________ Freesurfer mailing >>> list >>> Freesurfer@nmr.mgh.harvard.edu <mailto: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. >>> >>> _______________________________________________ >>> Freesurfer mailing list >>> Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> >>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >>> > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto: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
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@nmr.mgh.harvard.edu
-
Anderson Winkler -
jorge luis