ODU RESEARCHERS DEVELOP APPLICATION FOR NATIONAL COMPUTING GRID

Researchers at Old Dominion University have developed a computing application that can simulate an electric charge coursing through the body, and the payoff may be advances in treatments for disease or pain.

The Bioelectric Simulator for Whole Body Tissues (BioSim), as the application is formally known, models mammalian bio-response to electrical stimulation. The project is the first of its kind and has evolved from the university's resources in bioelectrics, computer modeling and simulation, as well as from its affiliations with the Southeastern Universities Research Association (SURA) and the National LambdaRail (NLR).

BioSim has been accepted as an application on the SURAgrid computing network that connects about 30 universities and research facilities, not only in the southeastern United States, but also in the Southwest and on the West Coast. By means of a grid, computers at many far-flung schools and research laboratories can work together as one super-computing unit.

Ravindra Joshi, University Professor of electrical and computer engineering and a researcher at the university's Frank Reidy Research Center for Bioelectrics, said the project demonstrates how engineers can "add to understanding of biological systems by making models and making them more and more realistic."

In the case of BioSim, it is grid computing-with dozens of computer clusters working together on an application-that allows realistic simulations of the response of mammal bodies to electrical stimulation. Researchers hope new diagnostic and therapeutic tools will result from experiments in which electrical stimulation is introduced by electrodes on the skin. It is safer and more efficient to test electric current-stimulation strategies on computational models of whole bodies rather than on live bodies.

Joshi is working on the project with his doctoral student Ashutosh Mishra and two staff members with the Office of Computing and Communications Services (OCCS), Mike Sachon, assistant director of research computing, and Mahantesh Halappanavar, grid-computing systems engineer.

The project has cutting-edge significance for two reasons. As far as the ODU researchers know, the application they have produced is the first that can show how a whole body responds to electrical current introduced via electrodes on the skin. Also, the application has been developed to be computer-grid-enabled. This means that computers in the SURA network across the nation could run the application simultaneously and produce a realistic snapshot of a body's response to a particular bioelectric strategy.

"We need to know how much voltage and duration can be used and still be within safe limits," Joshi said. A body also reacts differently to electrical current depending on where the current is introduced-where the electrodes are placed-and this is a factor that can be explored with the simulations. One concern would be electrodes placed too close to the heart.

"Previous work in this field has been on localized tissues of small volume," explained Mishra. For example, a computational model of the liver has been used for some time in simulations. "We're expanding it to the whole body, and because of this the variability greatly increases."

The variability comes from the variety of tissues, organs and other components in a whole body model. Every component may react differently as the charge plays out, say from a point of introduction in the shoulder, then through the trunk and into the lower extremities.

To picture the researchers' model, imagine a finely drawn, three-dimensional mesh. The mesh represents a map of millions of reference points within the whole body. More reference points means a finer detailed mesh and a more accurate simulation, but this also requires more computing power.

Mishra said a simulation involving a model of the human head would require 8.2 million reference points in order to provide acceptable accuracy. The whole-body model requires many more millions of points.

"Processing a data set of millions of mesh points creates prohibitively large requirements for system memory and computation times," says an ODU document about BioSim. "To address this problem, ODU researchers have demonstrated that BioSim scales well with the increased computational resources of distributed systems such as high-performance compute (HPC) clusters. By grid-enabling BioSim and deploying it in the SURAgrid environment, ODU's researchers have gained access to the HPC clusters that are part of SURAgrid's collective resources and have increased their ability to process the mesh point data in their large, whole-body model simulations."

ODU's vice president for research, Mohammad Karim said, "Professor Joshi's BioSim project is a perfect example of an application-driven design for a large-scale, multi-purpose grid infrastructure. This effort puts the work of outstanding ODU researchers to effective use by all the grid member institutions."

Sachon pointed out that the two-year-old SURAgrid is a "grassroots" effort to give ODU and its partners-including North Carolina State, Louisiana State, Tulane, Clemson, University of Kentucky and Texas Tech-an up-to-date computing grid.

Several other grids in the country were built from the ground up with large grants from research funding agencies. SURAgrid, on the other hand, is making the most of the in-house computing resources that were already in place at member institutions. Because the resources that support SURAgrid were not developed initially to be part of a grid, the grid architects must solve problems raised by a heterogeneous collection of hardware, software and software licensing agreements, according to Sachon and Halappanavar.

The problem solving is ongoing, and the engineers said they have not yet demonstrated the full capabilities of BioSim on SURAgrid. Nevertheless, the ODU team has been "going to school," as Halappanavar put it, by engineering a gateway allowing ODU computing resources to be put to use on SURAgrid for storm surge research at the University of North Carolina. "We learned so much by doing that," said Halappanavar, who added that the status of grid computing today reminds him of where electricity stood in the mid-19th century and where the Internet stood less than two decades ago.

The ODU team wrote the BioSim application using only open-source software, which means there will be no licensing hurdles that could prevent the application from running on computers at multiple institutions.

ODU was connected in 2006 to the National LambdaRail, a high-speed, fiber-optics network 100 times faster than the university's conventional Internet connection. It can boost the performance of SURAgrid and of the BioSim application because many of the SURA members have NLR access, Sachon said.