Application Of GPU-based Graphics Rendering And Parallel Computing In Optical Molecular Imaging

While optical molecular imaging, as one of the research focuses in molecular imaging, involves two algorithm studies: forward problem and inverse problem. The essential of forward problem is the particle transport problem, which is equivalent to solving the Boltzmann equation. Since it's not easy to work out the analytical solution, a random statistical method, like Monte Carlo method, is often used for an accurate numerical solution.To ensure the accuracy of Monte Carlo simulation, it is necessary to track a huge number of photons, which will increase the simulation time in proportion. In this paper, a parallel method based on GPU-HPC is presented, it will parallelized Monte Carlo algorithm of forward problem. Since restricted by the traditional hardwired, GPU is not efficient in parallel computation. While CUDA released by NVIDIA Corporation has higher programming performance, wider application area and supports better CUDA hardware performance compared with traditional GPU. Through parallelizing Monte Carlo algorithm of forward problem using CUDA, simulation results in this paper show good performance with high accuracy and in short simulated time.Meanwhile, to enhance visual effects of graphics render, this paper uses OpenGL for the GPU hardware-accelerated. OpenGL could enhance visual effects and quicken 3D graphics render, and facilitate the experimental observation of the simulation results.