An Effective CUDA Parallelization of Projection in Iterative Tomography Reconstruction.

Lizhe Xie,Bin Yan,Yining Hu,Huazhong Shu,Libo Zhang,Lin Wang,Benqiang Yang,Yang Chen,Limin Luo,Wenyuan Liu

doi:10.1371/journal.pone.0142184

Lizhe Xie, Bin Yan + Show 8 more

Open Access

PDF Available

https://doi.org/10.1371/journal.pone.0142184

Copy DOI

Export

Save

Cite

Journal: PloS one	Publication Date: Nov 30, 2015
Citations: 11	License type: CC BY 4.0

Affiliation: Nanjing Medical University

Abstract
Highlights/Summary
Full-Text PDF
Similar Papers

Abstract

Listen

Projection and back-projection are the most computationally intensive parts in Computed Tomography (CT) reconstruction, and are essential to acceleration of CT reconstruction algorithms. Compared to back-projection, parallelization efficiency in projection is highly limited by racing condition and thread unsynchronization. In this paper, a strategy of Fixed Sampling Number Projection (FSNP) is proposed to ensure the operation synchronization in the ray-driven projection with Graphical Processing Unit (GPU). Texture fetching is also used utilized to further accelerate the interpolations in both projection and back-projection. We validate the performance of this FSNP approach using both simulated and real cone-beam CT data. Experimental results show that compare to the conventional approach, the proposed FSNP method together with texture fetching is 10~16 times faster than the conventional approach based on global memory, and thus leads to more efficient iterative algorithm in CT reconstruction.

Highlights

Computed tomography (CT) has become one of the most widely used non-invasive medical imaging systems
This paper proposes an effective parallelization scheme Fixed Sampling Number Projection (FSNP) for the projection in iterative CT reconstruction algorithms
In this FSNP method, the sampling point number on each projection ray is fixed to ensure the synchronization of parallel computing

Summary

Introduction

Computed tomography (CT) has become one of the most widely used non-invasive medical imaging systems. As the rapid development of multi-slice CT, 3-D CT has replaced the 2-D CT in radiology routines by providing fast 3-D scanning. Extensive studies have demonstrated that iterative methods, based on an accurate system model, are capable of providing better reconstruction quality than analytical methods, especially under low-dose CT scans [1,2,3,4,5,6,7,8,9,10]. Due to the high computation cost in iterative reconstructions, FBP (Filtered Backprojection) based analytical reconstructions still take the main horsepower in current clinical reconstruction for 3-D CT [11].

Methods

Discussion

Conclusion