SCIRun User Mailing List

Re: [SCIRUN-USERS] libraries and 64bit computing

• From: David Weinstein <dmw@sci.utah.edu>
• Cc: <scirun-users@sci.utah.edu>
• Subject: Re: [SCIRUN-USERS] libraries and 64bit computing
• Date: Mon, 16 Feb 2004 14:20:19 -0700

Hi Darren,

Unfortunately, I don't know very much about EIT -- but I do know there's a wealth of literature out there on this topic.

I have implemented a simple network in BioPSE to solve what I call the parameterized EIT problem -- the net is in SCIRun/src/Packages/BioPSE/nets/eit-brain-slab.net Tom Ferree and I actually put together the first version of this net about six years ago. The approach here is to dramatically reduce the number of degrees of freedom in the EIT problem by assuming that we have an a priori segmentation of the volume into elements with distinct material types (e.g. skin, bone, CSF, gray matter, white matter). As input to the algorithm I assume a set of measured scalp potentials due to a known pair of current source/sink electrodes. The algorithm "guesses" the conductivity for each of the segmented materials (I specify an expected value, as well as variance and min/max from the literature) -- this guess provides the conductivity values for the forward model, we run a forward simulation based on the known source/sink, and then the resulting simulated output potentials are compared to the measured potentials in order to calculate a "misfit". The set of "guessed" conductivities and the "misfit" are fed into a simplex-search optimization algorithm, which iteratively chooses new conductivity sets in an effort to drive the misfit to zero.

This approach doesn't take into account anisotropy, though presumably if you could parameterize your anisotropy into a couple of degrees of freedom (e.g. you knew the primary fiber orientation from DTI and just wanted to recover two scalars representing the radial and tangential scales), then the algorithm could be easily extended to incorporate that data.

Cheers,
Dave

On Feb 14, 2004, at 12:45 PM, Darren Weber wrote:

Dear David,

thanks and congratulations on some very nice development philosophy and
practices.

Morgan Hough has pointed out a few publications now using electrical
impedance tomography (EIT) to obtain conductivity estimates.  I've been
curious about EIT and diffusion tensor imaging (DTI, or diffusion weighted
imaging, DTI) to better define conductivities for cerebral tissues.  I
wonder if you would like to comment or advise, from your experiences in
defining inhomogenous/anisotropic conductivities?

Best, Darren

----- Original Message -----
From: "David Weinstein" <dmw@sci.utah.edu>
Cc: <scirun-users@sci.utah.edu>
Sent: Thursday, February 12, 2004 6:52 PM
Subject: Re: [SCIRUN-USERS] libraries and 64bit computing

Hi Darren,

We haven't yet run SCIRun on an Opteron, so I can't yet tell you
whether it will work or not.  However, we're about to get several
Opterons here within the next week, so soon we will begin officially
supporting it as one of our platforms.

The 64-bit platform that we do have experience with is SGI Irix.
SCIRun has worked with Irix from the beginning, and performs quite
well.

In response to your FEM question: yes, SCIRun/BioPSE does support fully
inhomogeneous, anisotropic conductivities for our FEM and FDM
simulations -- you can build an Nx6 matrix that provides a unique
tensor for every element in your domain.  Please see the Appendix A of
the BioFEM tutorial for more information about getting finite element
models imported into BioPSE:
http://software.sci.utah.edu/doc/User/Tutorials/BioFEM/
BioFEM.html#appendix_a

Cheers,
Dave

On Feb 11, 2004, at 9:49 PM, Darren Weber wrote:

Hi,

novice questions, mainly directed at FEM computations on 64bit systems.

What 64bit hardware/OS do you recommend? Have you worked with Opteron
processors?

What matrix libraries does SCIRUN use?  What libraries for GUI
displays is
used?  Will SCIRUN compile on 64bit Opteron systems?

Does SCIRUN handle anisotropy in FEM/FDM forward calculations?

Best, Darren

--
Darren L. Weber, Postdoctoral Research Fellow,
Room 371-10, China Basin Landing, Dynamic Neuroimaging Laboratory,
Department of Radiology, University of California, San Francisco
185 Berry Street, San Francisco, CA 94143-0946

185 Berry Street. Suite 350
San Francisco, CA 94107

voice:(1) 415 353-9444
fax:  (1) 415 353-9421
www:   http://dnl.ucsf.edu/users/dweber

"People travel to wonder at the height of the mountains,
at the huge waves of the seas, at the long course of the rivers,
at the vast compass of the ocean, at the circular motion of the stars,
and yet they pass by themselves without wondering."

St.
Augustine,
354 430

===================================================================== ==
====
== The SCIRun Users mailing list: send email to majordomo@sci.utah.edu
==
== for more details.
==
== Please acknowledge use of SCIRun in your papers and reports:
==
==   see http://software.sci.utah.edu/scirun-biopse_citation.bib
==
===================================================================== ==
====

===========================================================================
== The SCIRun Users mailing list: send email to majordomo@sci.utah.edu   ==
== for more details.                                                     ==
== Please acknowledge use of SCIRun in your papers and reports:          ==
==   see http://software.sci.utah.edu/scirun-biopse_citation.bib         ==
===========================================================================