Abstract
We present and evaluate an algorithm for image reconstruction from a small number of projections in 3D x-ray computed tomography (CT). The proposed algorithm is similar to the class of projected gradient methods. Because each iteration of these algorithms for large 3D CT reconstruction is very computationally demanding, our goal is to devise an algorithm with fast convergence. To achieve this goal, in the proposed algorithm the gradient descent for reducing the measurement misfit term is carried out using a stochastic gradient descent iteration and the gradient directions are weighted using weights suggested by parallel coordinate descent. To further improve the speed of the algorithm, at each iteration we minimize the cost function on the subspace spanned by the direction of the current projected gradient and several previous update directions. We apply the proposed algorithm on simulated and real cone-beam projections and compare it with a well-known accelerated projected gradient algorithm, Monotone Fast Iterative Shrinkage-Thresholding Algorithm (MFISTA). Evaluations show that the rate of convergence of the proposed algorithm is superior to that of MFISTA
Access provided by Autonomous University of Puebla. Download to read the full chapter text
Chapter PDF
Similar content being viewed by others
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2015 Springer International Publishing Switzerland
About this paper
Cite this paper
Karimi, D., Ward, R., Ford, N. (2015). A weighted stochastic gradient descent algorithm for image reconstruction in 3D computed tomography. In: Jaffray, D. (eds) World Congress on Medical Physics and Biomedical Engineering, June 7-12, 2015, Toronto, Canada. IFMBE Proceedings, vol 51. Springer, Cham. https://doi.org/10.1007/978-3-319-19387-8_18
Download citation
DOI: https://doi.org/10.1007/978-3-319-19387-8_18
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-19386-1
Online ISBN: 978-3-319-19387-8
eBook Packages: EngineeringEngineering (R0)