Well, I have recently seen data where ICA/DR would not pick up midbrain contribution to one of the ICs of interest - presumably because of registration limits. But when going back to seeding the individual sessions based on what group Melodic had found you could demonstrate such contribution, yes. But it's something I have been involved with just recently. And I do agree, you must have good reasons to i) avoid circularity and ii) to not just show what you have already found. For example, look at all the signal plots often shown in model-driven FMRI: in a way, showing the blobs, COPE values in a table and % signal changes also tends to be quite redundant...
Cheers,
Andreas

________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Cornelius Werner [[log in to unmask]]
Gesendet: Mittwoch, 23. Februar 2011 13:13
An: [log in to unmask]
Betreff: Re: [FSL] AW: [FSL] seed based analysis

Good points. I'll keep that in mind. Do you have experience on that?
Cheers,
Cornelius

On Wed, Feb 23, 2011 at 11:41 AM, Andreas Bartsch
<[log in to unmask]> wrote:
> Hi,
>
> a reason I could think of is when the structure-of-interest is not well aligned (e.g. in the brainstem) and thus not well picked up by the ICA/DR (compared to other cortical structures) and to take advantage of the analyses at the single-subject level. In that regards, I think ICA/DR and single-subject seed-based analysis can be complementary / confirmatory under certain circumstances. But yes, the former may usually largely stand on its own.
> Cheers-
> Andreas
>
> ________________________________________
> Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Cornelius Werner [[log in to unmask]]
> Gesendet: Mittwoch, 23. Februar 2011 10:46
> An: [log in to unmask]
> Betreff: Re: [FSL] seed based analysis
>
> Hi.
>
> First of all: why would you want to do this? ICA is constructed such
> that the need for the selection of a (potentially biased) seed voxel
> is unnecessary. In terms of "seed" and "target", the ICA equals the
> seed, and the DR output basically equals the target. What is your aim
> doing the same on a seed based method? I am not sure that you aren't
> following circular logic here!
>
> Admittedly, before the advent of DR, one could have chosen (and still
> can choose) an interesting ICA component and define a ROI based on
> that for, e.g., a PPI analysis. Doing a seed based resting state (is
> that so?) analysis would include not only defining your ROI of
> interest, but also ROIs of NO interest, such as CSF and white matter,
> to correct for their (noisy) contributions, which basically means a
> full segmentation of your brains with all what that conveys. Ideally,
> you would also have other biosignals such as heartbeat and respiration
> as regressors of no interest  -something you (ideally) do not need in
> Melodic/DR, as those signals get their own components. Seed based
> analyses are tricky!
>
> But to answer at least one of your questions: yes, there are tools for
> signal extraction from a ROI, such as featquery. There have been
> several posts on that. As this is a little laborious, I'd suggest you
> search the archives on that if you are willing to go that way. Search
> for "define ROI from activation" and "featquery" and so forth.
>
> Good luck,
> Cornelius (from not-so-far Aachen)
>
> On Tue, Feb 22, 2011 at 7:50 PM, Perri Carol (di) (PSYCHOLOGY)
> <[log in to unmask]> wrote:
>> Dear FSL group, I have 2 questions for you, I will try to be as clear as possible.
>>
>> I have run melodic and dual regression on a group of patients and controls.
>> I got 25 components, some of them differing spacially on dual regression between the two groups.
>>
>> At this point I would like to study this components better with a seed-based analysis.
>>
>> Here come my questions.
>>
>> 1) Is there an easy way that I can select the voxel / ROI with the highest peak of activation both on the ICA components and on the dual regression output (dr_stage3_ic.._tfce_corrp_tstat..)?
>>
>>
>> 2)how do I do a seed based study? to be more clear how can I select a roi and extract time courses (means of ROI etc.) for post processing ?
>>
>> Thanks in advance for your help!
>>
>> Carol
>
>
>
> --
> Dr. med. Cornelius J. Werner
>
> Department of Neurology
> RWTH Aachen University
> Pauwelsstr. 30
> 52074 Aachen
> Germany
>
> Institute of Neuroscience and Medicine
> MR Physics - INM4
> Research Centre Juelich
> 52425 Juelich
> Germany
>
> ::: Please encrypt confidential data :::



--
Dr. med. Cornelius J. Werner

Department of Neurology
RWTH Aachen University
Pauwelsstr. 30
52074 Aachen
Germany

Institute of Neuroscience and Medicine
MR Physics - INM4
Research Centre Juelich
52425 Juelich
Germany

::: Please encrypt confidential data :::