Share Email Print
cover

Proceedings Paper

Implementation on GPU-based acceleration of the m-line reconstruction algorithm for circle-plus-line trajectory computed tomography
Author(s): Zengguang Li; Xiaoqi Xi; Yu Han; Bin Yan; Lei Li
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

The circle-plus-line trajectory satisfies the exact reconstruction data sufficiency condition, which can be applied in C-arm X-ray Computed Tomography (CT) system to increase reconstruction image quality in a large cone angle. The m-line reconstruction algorithm is adopted for this trajectory. The selection of the direction of m-lines is quite flexible and the m-line algorithm needs less data for accurate reconstruction compared with FDK-type algorithms. However, the computation complexity of the algorithm is very large to obtain efficient serial processing calculations. The reconstruction speed has become an important issue which limits its practical applications. Therefore, the acceleration of the algorithm has great meanings. Compared with other hardware accelerations, the graphics processing unit (GPU) has become the mainstream in the CT image reconstruction. GPU acceleration has achieved a better acceleration effect in FDK-type algorithms. But the implementation of the m-line algorithm’s acceleration for the circle-plus-line trajectory is different from the FDK algorithm. The parallelism of the circular-plus-line algorithm needs to be analyzed to design the appropriate acceleration strategy. The implementation can be divided into the following steps. First, selecting m-lines to cover the entire object to be rebuilt; second, calculating differentiated back projection of the point on the m-lines; third, performing Hilbert filtering along the m-line direction; finally, the m-line reconstruction results need to be three-dimensional-resembled and then obtain the Cartesian coordinate reconstruction results. In this paper, we design the reasonable GPU acceleration strategies for each step to improve the reconstruction speed as much as possible. The main contribution is to design an appropriate acceleration strategy for the circle-plus-line trajectory m-line reconstruction algorithm. Sheep-Logan phantom is used to simulate the experiment on a single K20 GPU. The development environment trajectory, using CPU and the paper’s GPU acceleration strategy, respectively. The experimental results show considerable reconstruction image quality, and the reconstruction acceleration ratio can reach 620 times.

Paper Details

Date Published: 25 October 2016
PDF: 8 pages
Proc. SPIE 10156, Hyperspectral Remote Sensing Applications and Environmental Monitoring and Safety Testing Technology, 101561Z (25 October 2016); doi: 10.1117/12.2247555
Show Author Affiliations
Zengguang Li, National Digital Switching System Engineering and Technological Research Ctr. (China)
Xiaoqi Xi, National Digital Switching System Engineering and Technological Research Ctr. (China)
Yu Han, National Digital Switching System Engineering and Technological Research Ctr. (China)
Bin Yan, National Digital Switching System Engineering and Technological Research Ctr. (China)
Lei Li, National Digital Switching System Engineering and Technological Research Ctr. (China)


Published in SPIE Proceedings Vol. 10156:
Hyperspectral Remote Sensing Applications and Environmental Monitoring and Safety Testing Technology

© SPIE. Terms of Use
Back to Top