Parallel Computing Lab
As the performancecost ratio of the personal computer (PC) and widebandwidth network technology rapidly increases,parallel computing with PC clusters has been widely recognized as an efficient method for high performance supercomputing, which has a long history. The parallel computing lab within the Biomedical Laser Laboratory was established in 2000 and has provided a worldclass research environment at East Carolina University for modeling lighttissue and lightcell interactions for researchers and students. Our current simulation research projects can be divided into two categories according to the underlying theoretical models. The first type of simulations is based on a statistical modeling of light transportation in biological tissues or, in general, turbid media. In this approach light is treated as a collection of particles (photons) without considering their phase information. The statistical approach involves treating light scattering and absorption as random events and the fate of a particular photon is determined randomly, as if rolling a dice in a Casino. This approach of simulation has been, properly and playfully, named as Monte Carlo Simulations. This simulation technique has been extensively used in physics, chemistry, biology, medicine and other fields for its simplicity in algorithm and implementation. Since the photons in the Monte Carlo simulation are treated independently, the algorithms can be made into parallel codes in a "natural" way by dividing tracked photons into many parcels to be distributed to each processing element of the cluster. The current focus of parallel Monte Carlo simulations is on the inverse determination of tissue optics parameters from imagebased reflectance measurements and modeling of light distribution in photodynamic therapy. The second type of our simulations is directed towards the modeling of the light interaction with single biological cell or cell spheroid. In this approach, we take into consideration the phase information of the light field by solving directly the Maxwell equations of classical electrodynamics to understand the microscopic origin of light scattering in biological tissues and to model light scatterig by a cell. Since these simulations require solving the differential equations with a finitedifferencetimedomain (FDTD) method, adapting this type of algorithms efficiently into parallel code is a much more challenging job. We completed initial development and testing of a parallel FDTD code in 2004 for modeling of light scattering by deformed red blood cells. Continued modeling of light scattering by B and HL60 cells is currently in progress with an automated 3D reconstruction of cell structure from confocal image stacks. The active research at the Biomedical Laser Laboratory has the unique feature of combining both experimental and theoretical investigations to study the challenging problems of tissue and cell optics. After your virtual tour of this web site, we would be more than glad to hear your questions or comments. Happy computing!
Jun Q. Lu, Ph.D., luj@ecu.edu 
