tsliu Wrote:
-------------------------------------------------------
> Hi Ziad,
> Thanks for your reply. I didn't know all this is
> possible, that's great news. I've seen the
> suma.pdf document. I thought it was lecture
> slides, but I guess it also serves the purpose of
> a manual? If so we shall study it carefully.
It is meant for the bootcamp, but if you install the bootcamp data it also acts as a manual of sorts. The slides that have 'hands on' printed in blue are meant to walk you through various aspects of the interface.
>
> I have a basic question about the general
> approach here. A colleague here who's helping me
> with this suggested (for my 2nd question, i.e.,
> group analysis) to 1) use AFNI to calculate %
> signal change in the native space, and then 2)
> warp those images to FreeSurfer and use
> FreeSurfer's statistical tools to do group
> analysis. In this method there's no need to use
> SUMA.
That would be one option. The process using SUMA still requires that you use FreeSurfer to create the surfaces and warp them to the spherical template (sphere.reg) but the way the SUMA pipeline is designed you will no longer need to warp the %signal change after they are computed. With AFNI/SUMA, you would compute the % signal change on standard mesh surfaces that were derived from the warped FreeSurfer surfaces. See the latest description of the process in this paper
You should also consider computing %signal change etc. AFTER you have mapped the time series onto the surface and smoothed those time series on the surface, rather than in the volume. afni_proc.py is setup to do this automatically for you. See afni_proc.py's example #8 for details. The data for that example is also from the bootcamp material.
>
> I actually don't know whether his suggested
> approach is viable, do you know?
Quite as I mentioned above, but perhaps a little less than optimal if blurring is done in the volume. Either way you can do the analysis quite readily.
> At least your way
> (using SUMA) is well-established, right?
Yes, it has been in use since 2006. The standard mesh approach has been used by Caret for creating the PALS atlas too.
> Another
> thing I'd like is the flexibility to use Caret's
> inter-subject alignment method, and from what I
> see, SUMA seems flexible on this aspect, in that
> it works with several registration framework,
> right?
Almost, in that I have not created standard-mesh surfaces from CARET's warps. However I suspect they can generate them within CARET, but you would have to ask John or Donna about that.

>Whereas I guess if we use FreeSurfer we can
> only use its own method--is this the right
> thinkgin?
You mean their warping approach? That is correct.

> Also do you happen to have some opinion
> on the pros and cons on Caret vs. FreeSurfer?
> --sorry if this is beyond the scope of this
> question ;)
Beyond the scope. I can tell you that FreeSurfer has the advantage of being fully automated, so for human data of decent quality, the process is quite painless. I do get CARET data from colleagues doing primate research and I hear that CARET is better suited for that task. For a more definite answer consider trying both!


>
> Please let me know. We need to make a major
> decision about what to do. We're thinking to
> attend the FreeSurfer workshop, but now if SUMA is
> the answer then we should attend the AFNI
> bootcamp--when is the next bootcamp, by the way?
> Thank you,
The next AFNI bootcamp will be sometime in mid to late spring at the earliest. Check the message board for updates on that. We just finished one two weeks ago and have not set a new date yet.
I'd say give the process a test drive then decide what you need to attend.

>
> --ts
cheers,
Ziad