Dear List (+ Sam, Andreas, Alex, M.),
To make things clearer (?), here is a message I had written and forgot
to send after Sam Reyes' first reply. Since then, we have had a few
offline exchanges but I am still not sure of how Curry implements Loreta
as far as individual brains go.
*****
...
My (quick) comments were based on the assumption that M. de Leonni
(please, pitch in to provide more info) was using the classical Loreta
spherical implementation which can only (i.e., most safely) relate to
anatomy via a warp based on the electrodes' locations in both spherical
and anatomical space.
Regarding your own approach, I am not sure whether Curry's
implementation uses spherical calculations warped in and/or masked by a
standard - resp. subject dependent - brain (e.g., gray matter only), or
if it directly estimates Loreta in a standard - resp. subject dependent
- brain. These approaches all have various intrinsic/theoretical
pros-and-cons that would certainly be better discussed by Roberto
Pascual-Marqui, Loreta's inventor (Roberto, are you here?).
If a standard brain is used by Curry, it could be MNI's (see LoretaKey)
statistical atlas which relates directly to the Talairach space and to
the SPM templates (as already extensively discussed in the list; please
remember that LOReta stands for LOwRes!). In this case - and even if
another brain than MNI's is used as the Curry Loreta standard - the
normalization to SPM should only be performed once (standard to SPM),
preferentially with the actual standard brain tissue maps rather than
with the binary brain mask reconstructed from the 10mm spaced grid.
If a subject dependent mask is used in Curry's calculation, it is likely
based on the gray matter segmented from the anatomical MRI of that same
subject. In this case, the corresponding tissue maps (same remark as
before) could be used to normalize the results to SPM space.
...
*****
I hope these comments will help some more, and I am still glad to see
that things are picking up on EEG related information.
Cheers,
Luc
> -----Original Message-----
> From: sareyes [mailto:[log in to unmask]]
> Sent: Wednesday, February 13, 2002 6:38 PM
> To: Bidaut, Luc
> Cc: SPM
> Subject: Re: coregistration: loreta EEG and MRI
>
>
> I have hacked together a method to normalize LORETA current
> density maps
> form the Curry source localization package with MR and PET
> image as part
> of a project to do cross subject comparisons. It is not yet
> published,
> but you are free to use it. I would also like to hear what any other
> SPMers (especially John) may think of it.
>
> The LORETA output format we get is a list of regularly spaced (10mm
> lattice) points in the brain and the current magnitude at
> each point. I
> create a binary brain mask from the points in the brain.
> Then I use a
> linear only normalization to normalize the LORETA brain mask to the
> brain mask distributed with SPM. I then apply this normalization to
> each of the LORETA images.
>
> For your purposes, I would suspect that by segmenting the MR
> you could
> create a brain mask for each subject and normalize the
> subject's LORETA
> to their MR.
>
> I have some matlab and perl scripts to automate some of this
> if you are
> using Curry. Let me know if you want them.
>
> Hope it helps,
>
> ---sam
>
|