Hi Christopher,
I've used BET in this manner, with matlab. The last time I used BET to get
the vertices, it output the vertices, but the faces were difficult to
interpret. I tried to get a right-handed triangulation of the vertices, but
it was beyond my time and ability (I could get a triangulation, just wasn't
sure if it was right handed). You can find my function to read the output
vertices at http://eeg.sf.net and especially see:
http://eeg.sourceforge.net/eegdoc/eeg_toolbox/mesh_bet2matlab.html
Anyhow, I switched to using a shrink wrap on the BET brain volume. There is
a function called mesh_shrinkwrap:
http://eeg.sourceforge.net/eegdoc/eeg_toolbox/mesh_shrinkwrap.html
Since creating that function and using it successfully, I may have changed
the default metric of the toolbox from mm to meters and that broke the input
parameters to mesh_shrinkwrap, but the latest version of the toolbox should
have a revised mesh_shrinkwrap that handles meters OK. In any case, you
need to be careful getting the parameters right, it's not robust at present.
I also modified how it shrinks, from a slower recursive process to an
intensity search along an interpolated line from the origin out to the
vertices. The latter algorithm is fast, but maybe not so robust. I should
switch back to the former method and probably refine the cost function
(probably using the same cost function as BET, only flipped so it works from
low intensity to higher intensity). This shrinkwrap function was also used
to get the scalp surface and the outer skull surface, where the latter
volume was also given by BET. The function moves the vertices only along a
radial line from the vertex to the origin, vertices are not allowed to move
laterally. This means that all vertices from one shell (eg, scalp to outer
skull) are coincident along a line from the volume origin. I think this has
implications for optimal calculations in BEM. I understand that Steve etal
are working on a new version of BET for just these purposes. I'm sure it
will be a useful development for those of us interested in more than brain
segmentation. Actually, as you are using MEG, it is a little surprising
that you are looking to tesselate the brain, all you need is the scalp. You
only need the skull (inner & outer) for EEG. In most cases, BEM treats the
brain and CSF as equivalent (conductivity). Anyhow, I've found that BET is
a very useful tool for this process and likely to get more so.
Take care, Darren
----- Original Message -----
From: "Christopher Joseph Bailey" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Monday, August 11, 2003 4:29 AM
Subject: Re: [FSL] BET: brain surface tesselation
> Dear All,
>
> I have another BET-related question/request. I have not yet delved into
> the source code to try to implement this myself.
>
> Would it be possible to add an output option for the final tesselated
> sphere (i.e. brain outline)? For example two files in Matlab-style:
> brain.nds (3D node positions) and brain.tri (indices into brain.nds
> forming triangles). A rough estimate of the brain outline would be
> useful---in my case for MEG boundary element modelling.
>
> Best regards,
>
> Christopher Bailey
> --
> Research Assistant (MSc)
> [log in to unmask]
> phn: +45 89 49 43 78
> fax: +45 89 49 44 00
>
> Center for Functionally Integrative Neuroscience (CFIN)
> Bygning 30, Trøjborgvej 72
> Aarhus Kommunehospital, DK-8000 Aarhus C
> Denmark
>
|