Hi Naama,
Steve would have a better take on this but he's away at the moment...
I think both approach are completely valid. The only problem is that you will not be able to compare the three groups altogether (in an F-test for instance) with a 2 groups registration approach (at least with the -S option).
What might happen is that the registration is maybe slightly less accurate for the 3 groups registration. Say you've got 3 groups: control, mild and severe groups. Then, grossly, a representative target will induce big warps for the control and severe groups and less for the mild group. On the other hand, if you choose a "2 group registration" approach, say between control and mild, the best target will be roughly midway between these two populations, with more moderate warping for the two of them...
If you're happy with what you've got at 0.05 corrected with the 3 groups approach, maybe you should stick to it (with one sole skeleton, it's easier to do all the possible analyses you might think of, including F-test).
Otherwise, you might want to create your own "study-specific, pediatric template" of FA by averaging the *_FA_to_target you've got so far, balancing for the n across the 3 groups, and then re-run the registration with this target specified (-t option). I've noticed it was working better for my elderly subjects (control, MCI and AD groups) than the -S or -T options...
Hope this helps!
Gwenaelle
--- En date de : Ven 1.5.09, Naama Barnea-Goraly <[log in to unmask]> a écrit :
> De: Naama Barnea-Goraly <[log in to unmask]>
> Objet: [FSL] Is this a valid thing to do? ( 3 groups tbss comparison)
> À: [log in to unmask]
> Date: Vendredi 1 Mai 2009, 0h11
> Follow up on this question - upon careful inspection it
> seems that the results of the individual t-tests run on two
> group registration and results of the t-tests run on the 3
> group registration are a little different.
> Essentially the clusters are slightly less significant in
> the (2 group) t-tests done with the skeleton derived from
> the 3 group registration such that if Iook at the results at
> 0.05 in the 3 groups they will look the same as at 0.01 at
> the 2 groups registration. If I increase the threshold to
> 0.01 at the 3 group registration some smaller clusters will
> be lost. So my question is which results are valid and
> should be reported?
>
> Thanks,
>
> Naama
>
>
>
> On Apr 21, 2009, at 9:51 AM, Gwenaëlle DOUAUD wrote:
>
> > Hi Naama,
> >
> > yes you're right, in this case, this is better to
> run tbss3 with the -S option...
> >
> > Cheers,
> > Gwenaëlle
> >
> >
> > --- En date de : Mar 21.4.09, Naama Barnea-Goraly
> <[log in to unmask]> a écrit :
> >
> >> De: Naama Barnea-Goraly
> <[log in to unmask]>
> >> Objet: Re: [FSL] Re : [FSL] Is this a valid thing
> to do? ( 3 groups tbss comparison)
> >> À: [log in to unmask]
> >> Date: Mardi 21 Avril 2009, 18h36
> >> Thanks Gwenaelle,
> >>
> >> This is a pediatric sample, so I assume it is
> better to run
> >> with the -S option ? Not a problem, I already ran
> it and the
> >> skeletons are all the same now, was just checking
> ;-)
> >>
> >> Best,
> >>
> >> Naama
> >>
> >> On Apr 21, 2009, at 9:23 AM, Gwenaëlle DOUAUD
> wrote:
> >>
> >>> Hi Naama,
> >>>
> >>> yes, it would be valid, but I presume you
> don't
> >> need to rerun the whole 2-4 steps...
> >>> It looks slightly "misregistered"
> quite
> >> likely because you ran tbss3 with the -S option,
> so the mean
> >> FA and skeleton are derived for each of your
> group. The
> >> skeletons are different, but the registration
> should be the
> >> same if you have run the tbss2 step with the
> FMRIB58
> >> template.
> >>> If so, then you don't need to run it
> again. You
> >> just need to run the tbss3 with the option -T with
> all your
> >> subjects in the same FA directory, then tbss4.
> Basically
> >> what you want here is a common skeleton in order
> to be able
> >> to compare each group.
> >>>
> >>> Cheers,
> >>> Gwenaëlle
> >>>
> >>>
> >>> --- En date de : Mar 21.4.09, Naama
> Barnea-Goraly
> >> <[log in to unmask]> a écrit :
> >>>
> >>>> De: Naama Barnea-Goraly
> >> <[log in to unmask]>
> >>>> Objet: [FSL] Is this a valid thing to do?
> ( 3
> >> groups tbss comparison)
> >>>> À: [log in to unmask]
> >>>> Date: Mardi 21 Avril 2009, 17h48
> >>>> I have 3 groups: subjects, relatives and
> controls.
> >> We ran
> >>>> separate t-tests to get our results, but
> now I
> >> would like to
> >>>> show them on one image and when I overlay
> them
> >> they are
> >>>> slightly "misregistered". I am
> assuming
> >> this is
> >>>> because the registration is a little
> different in
> >> each
> >>>> comparison. I would also like to extract
> the
> >> values from
> >>>> regions of significant between group
> differences.
> >>>>
> >>>> Is it valid to run tbss_2- tbss_4 on all
> subjects
> >> (all 3
> >>>> groups together) and then fslsplit/fslroi
> the
> >> groups that I
> >>>> want, run analysis, generate images and
> extract
> >> values for
> >>>> graphs?
> >>>>
> >>>> Thanks,
> >>>>
> >>>> Naama
> >>>>
> >>>>
> >>>>
> >>>> Naama Barnea-Goraly M.D.
> >>>> Instructor
> >>>> Center for Interdisciplinary Brain
> Sciences
> >> Research
> >>>> Stanford University Division of Child and
> >> Adolescent
> >>>> Psychiatry
> >>>> 401 Quarry Rd. MC 5795
> >>>> Stanford University School of Medicine
> >>>> Stanford, CA 94305-5795
> >>>> Phone: (650) 736-1874, fax: (650) 724-4794
> >>>>
> >>>>
> >>>> CONFIDENTIALITY NOTICE: This e-mail
> communication
> >> and any
> >>>> attachments may contain confidential
> information
> >> for the use
> >>>> of the designated recipients named above.
> If you
> >> are not
> >>>> the intended recipient, you are hereby
> notified
> >> that you
> >>>> have received this communication in error
> and that
> >> any
> >>>> review, disclosure, dissemination,
> distribution or
> >> copying
> >>>> of it or its contents is prohibited. If
> you have
> >> received
> >>>> this communication in error, please notify
> >> Stanford Medical
> >>>> Center immediately by telephone at (650)
> 725-5722
> >> and
> >>>> destroy all copies of this communication
> and any
> >>>> attachments. Thank you.
> >>>
> >>>
> >>>
> >>
> >> Naama Barnea-Goraly M.D.
> >> Instructor
> >> Center for Interdisciplinary Brain Sciences
> Research
> >> Stanford University Division of Child and
> Adolescent
> >> Psychiatry
> >> 401 Quarry Rd. MC 5795
> >> Stanford University School of Medicine
> >> Stanford, CA 94305-5795
> >> Phone: (650) 736-1874, fax: (650) 724-4794
> >>
> >>
> >> CONFIDENTIALITY NOTICE: This e-mail communication
> and any
> >> attachments may contain confidential information
> for the use
> >> of the designated recipients named above. If you
> are not
> >> the intended recipient, you are hereby notified
> that you
> >> have received this communication in error and that
> any
> >> review, disclosure, dissemination, distribution or
> copying
> >> of it or its contents is prohibited. If you have
> received
> >> this communication in error, please notify
> Stanford Medical
> >> Center immediately by telephone at (650) 725-5722
> and
> >> destroy all copies of this communication and any
> >> attachments. Thank you.
> >
> >
> >
>
> Naama Barnea-Goraly M.D.
> Instructor
> Center for Interdisciplinary Brain Sciences Research
> Stanford University Division of Child and Adolescent
> Psychiatry
> 401 Quarry Rd. MC 5795
> Stanford University School of Medicine
> Stanford, CA 94305-5795
> Phone: (650) 736-1874, fax: (650) 724-4794
>
>
> CONFIDENTIALITY NOTICE: This e-mail communication and any
> attachments may contain confidential information for the use
> of the designated recipients named above. If you are not
> the intended recipient, you are hereby notified that you
> have received this communication in error and that any
> review, disclosure, dissemination, distribution or copying
> of it or its contents is prohibited. If you have received
> this communication in error, please notify Stanford Medical
> Center immediately by telephone at (650) 725-5722 and
> destroy all copies of this communication and any
> attachments. Thank you.
|