Ultrafast CT Reconstruction on the NVIDIA GeForce 8800

This thesis describes the development of an ultrafast computed tomography (CT) reconstruction engine, coupling fast CT reconstruction algorithms with the parallelism and processing power of the NVIDIA GeForce 8800 graphics processing unit (GPU). InstaRecon, a local startup company, has developed advanced mathematical algorithms that make the reconstruction process more computationally efficient, providing speedups of 20 to 100-fold. The NVIDIA GeForce 8800 represents a new generation of GPUs, with a unified architecture that provides tremendous processing power through a high degree of parallelism. CUDA, which stands for Compute Unified Device Architecture, is a recently released development environment, facilitates the design and implementation of general purpose (nongraphics) algorithms on these new GPUs. The aim of this project is to combine the software acceleration of fast reconstruction algorithms with the hardware acceleration of this highly parallel platform, resulting in an ultrafast reconstruction engine at a low cost.unpublishednot peer reviewedU of I OnlyUndergraduate senior thesis not recommended for open acces