Frank Marshall and Tong Xu

Printer-friendly version
Date: 
Thursday, November 21, 2013
Time: 
3:30 pm
Location: 
Conference room A&B, Room C2362, 2nd floor, Cancer Centre South , The Ottawa Hospital (General Campus)

1. "Reconstruction of a Distributed Radioactive Source with a Directional Spectrometer"
Frank Marshall - Carleton University
Abstract: The Emergency Response Group at Natural Resources Canada is responsible for developing innovative techniques of reconstruction for localizing and mapping radioactive sources.  In one area of research, the group is has been in joint collaboration with Defense Research and Development Canada (DRDC) to determine novel techniques for mapping radioactive distributed sources (RDDs).  Over the past two years, Medicine Hat in Suffield, Alberta has provided the testing grounds for several experiments, in which lanthanum-140 sources were detonated.  In these experiments, a directional spectrometer was used to record the spatial variation of the source intensity.  It consists of four, tightly-packed, NaI detectors.  It was mounted on a truck and driven around the source distribution.  From this survey, the limited data of points along the truck path leave much information to be extracted regarding the true source distribution.  This talk will review some of the methods that are employed to approximate the local intensity in the vicinity of the trucks path.  In particular, there will be a review of the method used to determine a factor that converts the measured signal into an intensity measurement for the case of the detector overlying an infinite disc source.  This method makes use of a curve of counts versus disc radius, which is called the detector footprint.  Results of EGSnrc simulations will be presented for this calculation, as will results of detector parameter simulations.



2. "A GPU implementation of EGSnrc"
Tong Xu - Carleton University
Abstract: As an effort to enable accurate and fast Monte Carlo simulation for potential clinical use, the physics core of the well accepted Monte Carlo simulation package, EGSnrc, was implemented on the parallel computing platform based on GPU, Graphics Process Units. With hundreds of processors integrated in one cost effective board, GPU has recently shown great potential on high performance computing, including Monte Carlo simulations.  An introduction to the concept of GPU computing will be given. The simulation structure of EGSnrc was changed to achieve better performance on GPU. Through the simulation of PDDs and dose profiles, the newly developed GPU based system was benchmarked and validated against the original EGSnrc.