Al-Rifaie, Mohammad Majid ORCID: 0000-0002-1798-9615 and Blackwell, Tim (2022) Swarm led tomographic reconstruction. In: GECCO '22: Proceedings of the Genetic and Evolutionary Computation Conference. UNSPECIFIED, pp. 1121-1129. ISBN 9781450392372 (doi:https://doi.org/10.1145/3512290.3528737)

Preview

PDF 37116-AL-RIFAIE-Swarm led tomographic reconstruction.pdf - Accepted Version Download (1MB) | Preview

Abstract

Image reconstruction from ray projections is a common technique in medical imaging. In particular, the few-view scenario, in which the number of projections is very limited, is important for cases where the patient is vulnerable to potentially damaging radiation. This paper considers swarm-based reconstruction where individuals, or particles, swarm in image space in an attempt to lower the reconstruction error. We compare several swarm algorithms with standard algebraic reconstruction techniques and filtered backprojection for five standard test phantoms viewed under reduced projections. We find that although swarm algorithms do not produce solutions with lower reconstruction errors, they generally find more accurate reconstructions; that is, swarm techniques furnish reconstructions that are more similar to the original phantom. A function profiling method suggests that the ability of the swarm to optimise these high dimensional problems can be attributed to a broad funnel leading to complex structure close to the optima. This finding is further exploited by optimising the parameters of the best performing swarm technique, and the results are compared against three unconstrained and boxed local search methods. The tomographic reconstruction-optimised swarm technique is shown to be superior to prominent algebraic reconstructions and local search algorithms.

Item Type:	Conference Proceedings
Title of Proceedings:	GECCO '22: Proceedings of the Genetic and Evolutionary Computation Conference
Uncontrolled Keywords:	tomographic reconstruction, swarm optimisation, function profiling, dispersive flies optimisation
Subjects:	Q Science > Q Science (General)
Faculty / School / Research Centre / Research Group:	Faculty of Engineering & Science Faculty of Engineering & Science > School of Computing & Mathematical Sciences (CMS)
Related URLs:	Publisher
Last Modified:	05 Aug 2022 14:06
URI:	http://gala.gre.ac.uk/id/eprint/37116

Actions (login required)

View Item

Downloads