A Fast GPU-Based Approach to Branchless （一个快速的基于gpu的无枝的方法）.pdf

A Fast GPU-Based Approach to Branchless Distance-Driven Projection and Back-Projection in Cone Beam CT Daniel Schlifskeab and Henry Medeirosa aMarquette University, 1250 W Wisconsin Ave, Milwaukee, WI, USA; bGE Healthcare Imaging, 3000 N Grandview Blvd, Waukesha, WI, USA; ABSTRACT Modern image reconstruction algorithms rely on projection and back-projection operations to refine an image estimate in iterative image reconstruction. A widely-used state-of-the-art technique is distance-driven projection and back- projection. While the distance-driven technique yields superior image quality in iterative algorithms, it is a computationally demanding process. This has a detrimental effect on the relevance of the algorithms in clinical settings. A few methods have been proposed for enhancing the distance-driven technique in order to take advantage of modern computer hardware. This paper explores a two-dimensional extension of the branchless method proposed by Samit Basu and Bruno De Man. The extension of the branchless method is named 鈥減re-integration鈥 because it gets a performance boost by integrating the data before the projection and back-projection operations. It was written with NVIDIA鈥檚 CUDA platform and carefully designed for massively parallel GPUs. The performance and the image quality of the pre- integration method were analyzed. Both projection and back-projection are significantly faster with pre-integration. The image quality was analyzed using cone beam image reconstruction algorithms within Jeffrey Fessler鈥檚 Image Reconstruction Toolbox. Images produced from regularized, iterative image reconstruction algorithms using the pre- integration metho