Accelerated gradient methods for total-variation-based CT image reconstruction

Publication: Research - peer-reviewArticle in proceedings – Annual report year: 2011

Standard

Accelerated gradient methods for total-variation-based CT image reconstruction. / Jørgensen, Jakob Heide; Jensen, Tobias Lindstrøm; Hansen, Per Christian; Jensen, Søren Holdt; Sidky, Emil Y.; Pan, Xiaochuan.

Proceedings of 11th Fully 3D Meeting. 2011.

Publication: Research - peer-reviewArticle in proceedings – Annual report year: 2011

Harvard

Jørgensen, JH, Jensen, TL, Hansen, PC, Jensen, SH, Sidky, EY & Pan, X 2011, 'Accelerated gradient methods for total-variation-based CT image reconstruction'. in Proceedings of 11th Fully 3D Meeting.

APA

Jørgensen, J. H., Jensen, T. L., Hansen, P. C., Jensen, S. H., Sidky, E. Y., & Pan, X. (2011). Accelerated gradient methods for total-variation-based CT image reconstruction. In Proceedings of 11th Fully 3D Meeting.

CBE

Jørgensen JH, Jensen TL, Hansen PC, Jensen SH, Sidky EY, Pan X. 2011. Accelerated gradient methods for total-variation-based CT image reconstruction. In Proceedings of 11th Fully 3D Meeting.

MLA

Vancouver

Jørgensen JH, Jensen TL, Hansen PC, Jensen SH, Sidky EY, Pan X. Accelerated gradient methods for total-variation-based CT image reconstruction. In Proceedings of 11th Fully 3D Meeting. 2011.

Author

Jørgensen, Jakob Heide; Jensen, Tobias Lindstrøm; Hansen, Per Christian; Jensen, Søren Holdt; Sidky, Emil Y.; Pan, Xiaochuan / Accelerated gradient methods for total-variation-based CT image reconstruction.

Proceedings of 11th Fully 3D Meeting. 2011.

Publication: Research - peer-reviewArticle in proceedings – Annual report year: 2011

Bibtex

@inbook{b4c4bd8cfdb2495fb7a21b73a5fdd0ed,
title = "Accelerated gradient methods for total-variation-based CT image reconstruction",
author = "Jørgensen, {Jakob Heide} and Jensen, {Tobias Lindstrøm} and Hansen, {Per Christian} and Jensen, {Søren Holdt} and Sidky, {Emil Y.} and Xiaochuan Pan",
year = "2011",
booktitle = "Proceedings of 11th Fully 3D Meeting",

}

RIS

TY - GEN

T1 - Accelerated gradient methods for total-variation-based CT image reconstruction

A1 - Jørgensen,Jakob Heide

A1 - Jensen,Tobias Lindstrøm

A1 - Hansen,Per Christian

A1 - Jensen,Søren Holdt

A1 - Sidky,Emil Y.

A1 - Pan,Xiaochuan

AU - Jørgensen,Jakob Heide

AU - Jensen,Tobias Lindstrøm

AU - Hansen,Per Christian

AU - Jensen,Søren Holdt

AU - Sidky,Emil Y.

AU - Pan,Xiaochuan

PY - 2011

Y1 - 2011

N2 - Total-variation (TV)-based CT image reconstruction has shown experimentally to be capable of producing accurate reconstructions from sparse-view data. In particular TV-based reconstruction is very well suited for images with piecewise nearly constant regions. Computationally, however, TV-based reconstruction is much more demanding, especially for 3D imaging, and the reconstruction from clinical data sets is far from being close to real-time. This is undesirable from a clinical perspective, and thus there is an incentive to accelerate the solution of the underlying optimization problem. The TV reconstruction can in principle be found by any optimization method, but in practice the large scale of the systems arising in CT image reconstruction preclude the use of memory-demanding methods such as Newton’s method. The simple gradient method has much lower memory requirements, but exhibits slow convergence. In the present work we address the question of how to reduce the number of gradient method iterations needed to achieve a high-accuracy TV reconstruction. We consider the use of two accelerated gradient-based methods, GPBB and UPN, to solve the 3D-TV minimization problem in CT image reconstruction. The former incorporates several heuristics from the optimization literature such as Barzilai-Borwein (BB) step size selection and nonmonotone line search. The latter uses a cleverly chosen sequence of auxiliary points to achieve a better convergence rate. The methods are memory efficient and equipped with a stopping criterion to ensure that the TV reconstruction has indeed been found. An implementation of the methods (in C with interface to Matlab) is available for download from http://www2.imm.dtu.dk/pch/TVReg/. We compare the proposed methods with the standard gradient method, applied to a 3D test problem with synthetic few-view data. We find experimentally that for realistic parameters the proposed methods significantly outperform the gradient method.

AB - Total-variation (TV)-based CT image reconstruction has shown experimentally to be capable of producing accurate reconstructions from sparse-view data. In particular TV-based reconstruction is very well suited for images with piecewise nearly constant regions. Computationally, however, TV-based reconstruction is much more demanding, especially for 3D imaging, and the reconstruction from clinical data sets is far from being close to real-time. This is undesirable from a clinical perspective, and thus there is an incentive to accelerate the solution of the underlying optimization problem. The TV reconstruction can in principle be found by any optimization method, but in practice the large scale of the systems arising in CT image reconstruction preclude the use of memory-demanding methods such as Newton’s method. The simple gradient method has much lower memory requirements, but exhibits slow convergence. In the present work we address the question of how to reduce the number of gradient method iterations needed to achieve a high-accuracy TV reconstruction. We consider the use of two accelerated gradient-based methods, GPBB and UPN, to solve the 3D-TV minimization problem in CT image reconstruction. The former incorporates several heuristics from the optimization literature such as Barzilai-Borwein (BB) step size selection and nonmonotone line search. The latter uses a cleverly chosen sequence of auxiliary points to achieve a better convergence rate. The methods are memory efficient and equipped with a stopping criterion to ensure that the TV reconstruction has indeed been found. An implementation of the methods (in C with interface to Matlab) is available for download from http://www2.imm.dtu.dk/pch/TVReg/. We compare the proposed methods with the standard gradient method, applied to a 3D test problem with synthetic few-view data. We find experimentally that for realistic parameters the proposed methods significantly outperform the gradient method.

UR - http://arxiv.org/abs/1105.4002

UR - http://www.fully3d.org/2011/index.html

BT - Proceedings of 11th Fully 3D Meeting

T2 - Proceedings of 11th Fully 3D Meeting

ER -