Accelerating computed tomography on commodity graphics hardware

The task of reconstructing an object from its projections via tomographic methods is a time-consuming process due to the vast complexity of the data. For this reason, manufacturers of equipment for medical computed tomography (CT) rely mostly on special ASICs to obtain the fast reconstruction times required in clinical settings. Although modern CPUs have gained sufficient power in recent years to be competitive for 2D reconstruction, this is not the case for 3D reconstructions, especially not when iterative algorithms must be applied. The recent evolution of commodity programmable PC computer graphics boards (GPUs) has the potential to change this picture in a very dramatic way. In this thesis, we first show that many types of CT algorithms, both iterative and non-iterative, can greatly benefit from the high degree of SIMD (Same Instruction Multiple Data) parallelism these platforms provide. By doing so, results of high-fidelity can be obtained at speedups of over an order of magnitude. In addition to describing theories and implementation details, we further show dedicated solutions for resolving various challenges presented in cone-beam reconstruction using Feldkamp’s method on GPU. We also propose optimization techniques specifically targeting the latest GPU architecture that enables the implementation of our streaming-CT notion. Next, we use electronic microscopy tomography as an example to demonstrate the power of GPU’s computational capability which is even more important here due to EMT’s extensive usage of iterative algorithms. Here a sinogram-based method was designed to achieve the maximum speedup and system scalability. Last, a new rendering method D2VR that can produce higher visualization quality than traditional volume rendering algorithms but suffers from high computational complexity, is accelerated by our rendering-driven rapid-CT framework to obtain near-interactive framerates.