GPU Accelerated Structure-Exploiting Matched Forward and Back Projection for Algebraic Iterative Cone Beam CT Reconstruction

Thompson, William M and Lionheart, William RB (2014) GPU Accelerated Structure-Exploiting Matched Forward and Back Projection for Algebraic Iterative Cone Beam CT Reconstruction. The Third International Conference on Image Formation in X-Ray Computed Tomography, 22-25 June 2014, Salt Lake City, Utah, USA..

PDF CT_meeting_2014_final.pdf Download (2MB)

Abstract

Algebraic iterative reconstruction algorithms have been the subject of much research into problems of limited data CT reconstruction. However, their use for practical problems, particularly for high resolution systems and corresponding large volume sizes, has often been considered unfeasible due to their high computational demands. For cone beam geometry, we present a highly parallel adaptation of Siddon�s algorithm for discrete forward and back projection, based on exploitation of structure in the pattern of the cone beam rays. The proposed algorithm has been implemented for nVidia GPUs using CUDA, resulting in speedup of an order of magnitude when tested against a non-structure exploiting parallel CPU implementation of Ja- cobs� algorithm. The work is presented in the context of circular scan, flat panel detector micro CT, but could easily be adapted to other scanning trajectories and detector configurations.

Item Type:	Article
Uncontrolled Keywords:	CT reconstruction, GPU, algebraic iterative, CUDA
Subjects:	MSC 2010, the AMS's Mathematics Subject Classification > 47 Operator theory MSC 2010, the AMS's Mathematics Subject Classification > 65 Numerical analysis
Depositing User:	Prof WRB Lionheart
Date Deposited:	19 Sep 2016
Last Modified:	20 Oct 2017 14:13
URI:	https://eprints.maths.manchester.ac.uk/id/eprint/2319

Actions (login required)

View Item