Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL)

Ameet Kumar Jain, Yu Zhou, Tabish Mustufa, E. Clif Burdette, Gregory S. Chirikjian, Gabor Fichtinger

Research output: Contribution to journalArticle

Abstract

Intraoperative dosimetric quality assurance in prostate brachytherapy critically depends on discerning the three-dimensional (3D) locations of implanted seeds. The ability to reconstruct the implanted seeds intraoperatively will allow us to make immediate provisions for dosimetric deviations from the optimal implant plan. A method for seed reconstruction from segmented C-arm fluoroscopy images is proposed. The 3D coordinates of the implanted seeds can be calculated upon resolving the correspondence of seeds in multiple x-ray images. We formalize seed-matching as a combinatorial optimization problem, which has salient features: (a) extensively studied solutions by the computer science community; (b) proof for the nonexistence of any polynomial time exact algorithm; and (c) a practical pseudo-polynomial algorithm that mostly runs in O(N3) time using any number of images. We prove that two images are insufficient to correctly match the seeds, while a third image renders the matching problem to be of nonpolynomial complexity. We utilize the special structure of the problem and propose a pseudopolynomial time algorithm. Using three presegmented images, matching and reconstruction of brachytherapy seeds using the Hungarian algorithm achieved complete matching in simulation experiments; and 98.5% in phantom experiments. 3D reconstruction error for correctly matched seeds has a mean of 0.63 mm, and 0.9 mm for incorrectly matched seeds. The maximum seed reconstruction error in each implant was typically around 1.32 mm. Both on synthetic data and in phantom experiments, matching rate and reconstruction error achieved using presegmented images was found to be sufficient for prostate brachytherapy. The algorithm is extendable to deal with arbitrary number of images without any loss in speed or accuracy. The algorithm is sufficiently generic to provide a practical solution to any correspondence problem, across different imaging modalities and features.

Original languageEnglish (US)
Pages (from-to)3475-3492
Number of pages18
JournalMedical Physics
Volume32
Issue number11
DOIs
StatePublished - Nov 2005

Fingerprint

Brachytherapy
Seeds
Prostate
Computer-Assisted Image Processing
Fluoroscopy
X-Rays

Keywords

  • C-arm
  • Fluoroscopy
  • Prostate brachytherapy
  • Radiation planning
  • Reconstruction
  • Seed matching

ASJC Scopus subject areas

  • Biophysics

Cite this

Jain, A. K., Zhou, Y., Mustufa, T., Burdette, E. C., Chirikjian, G. S., & Fichtinger, G. (2005). Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL). Medical Physics, 32(11), 3475-3492. https://doi.org/10.1118/1.2104087

Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL). / Jain, Ameet Kumar; Zhou, Yu; Mustufa, Tabish; Burdette, E. Clif; Chirikjian, Gregory S.; Fichtinger, Gabor.

In: Medical Physics, Vol. 32, No. 11, 11.2005, p. 3475-3492.

Research output: Contribution to journalArticle

Jain, AK, Zhou, Y, Mustufa, T, Burdette, EC, Chirikjian, GS & Fichtinger, G 2005, 'Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL)', Medical Physics, vol. 32, no. 11, pp. 3475-3492. https://doi.org/10.1118/1.2104087
Jain AK, Zhou Y, Mustufa T, Burdette EC, Chirikjian GS, Fichtinger G. Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL). Medical Physics. 2005 Nov;32(11):3475-3492. https://doi.org/10.1118/1.2104087
Jain, Ameet Kumar ; Zhou, Yu ; Mustufa, Tabish ; Burdette, E. Clif ; Chirikjian, Gregory S. ; Fichtinger, Gabor. / Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL). In: Medical Physics. 2005 ; Vol. 32, No. 11. pp. 3475-3492.
@article{2805bd1decbd40a3b5756c256309d157,
title = "Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL)",
abstract = "Intraoperative dosimetric quality assurance in prostate brachytherapy critically depends on discerning the three-dimensional (3D) locations of implanted seeds. The ability to reconstruct the implanted seeds intraoperatively will allow us to make immediate provisions for dosimetric deviations from the optimal implant plan. A method for seed reconstruction from segmented C-arm fluoroscopy images is proposed. The 3D coordinates of the implanted seeds can be calculated upon resolving the correspondence of seeds in multiple x-ray images. We formalize seed-matching as a combinatorial optimization problem, which has salient features: (a) extensively studied solutions by the computer science community; (b) proof for the nonexistence of any polynomial time exact algorithm; and (c) a practical pseudo-polynomial algorithm that mostly runs in O(N3) time using any number of images. We prove that two images are insufficient to correctly match the seeds, while a third image renders the matching problem to be of nonpolynomial complexity. We utilize the special structure of the problem and propose a pseudopolynomial time algorithm. Using three presegmented images, matching and reconstruction of brachytherapy seeds using the Hungarian algorithm achieved complete matching in simulation experiments; and 98.5{\%} in phantom experiments. 3D reconstruction error for correctly matched seeds has a mean of 0.63 mm, and 0.9 mm for incorrectly matched seeds. The maximum seed reconstruction error in each implant was typically around 1.32 mm. Both on synthetic data and in phantom experiments, matching rate and reconstruction error achieved using presegmented images was found to be sufficient for prostate brachytherapy. The algorithm is extendable to deal with arbitrary number of images without any loss in speed or accuracy. The algorithm is sufficiently generic to provide a practical solution to any correspondence problem, across different imaging modalities and features.",
keywords = "C-arm, Fluoroscopy, Prostate brachytherapy, Radiation planning, Reconstruction, Seed matching",
author = "Jain, {Ameet Kumar} and Yu Zhou and Tabish Mustufa and Burdette, {E. Clif} and Chirikjian, {Gregory S.} and Gabor Fichtinger",
year = "2005",
month = "11",
doi = "10.1118/1.2104087",
language = "English (US)",
volume = "32",
pages = "3475--3492",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "11",

}

TY - JOUR

T1 - Matching and reconstruction of brachytherapy seeds using the Hungarian algorithm (MARSHAL)

AU - Jain, Ameet Kumar

AU - Zhou, Yu

AU - Mustufa, Tabish

AU - Burdette, E. Clif

AU - Chirikjian, Gregory S.

AU - Fichtinger, Gabor

PY - 2005/11

Y1 - 2005/11

N2 - Intraoperative dosimetric quality assurance in prostate brachytherapy critically depends on discerning the three-dimensional (3D) locations of implanted seeds. The ability to reconstruct the implanted seeds intraoperatively will allow us to make immediate provisions for dosimetric deviations from the optimal implant plan. A method for seed reconstruction from segmented C-arm fluoroscopy images is proposed. The 3D coordinates of the implanted seeds can be calculated upon resolving the correspondence of seeds in multiple x-ray images. We formalize seed-matching as a combinatorial optimization problem, which has salient features: (a) extensively studied solutions by the computer science community; (b) proof for the nonexistence of any polynomial time exact algorithm; and (c) a practical pseudo-polynomial algorithm that mostly runs in O(N3) time using any number of images. We prove that two images are insufficient to correctly match the seeds, while a third image renders the matching problem to be of nonpolynomial complexity. We utilize the special structure of the problem and propose a pseudopolynomial time algorithm. Using three presegmented images, matching and reconstruction of brachytherapy seeds using the Hungarian algorithm achieved complete matching in simulation experiments; and 98.5% in phantom experiments. 3D reconstruction error for correctly matched seeds has a mean of 0.63 mm, and 0.9 mm for incorrectly matched seeds. The maximum seed reconstruction error in each implant was typically around 1.32 mm. Both on synthetic data and in phantom experiments, matching rate and reconstruction error achieved using presegmented images was found to be sufficient for prostate brachytherapy. The algorithm is extendable to deal with arbitrary number of images without any loss in speed or accuracy. The algorithm is sufficiently generic to provide a practical solution to any correspondence problem, across different imaging modalities and features.

AB - Intraoperative dosimetric quality assurance in prostate brachytherapy critically depends on discerning the three-dimensional (3D) locations of implanted seeds. The ability to reconstruct the implanted seeds intraoperatively will allow us to make immediate provisions for dosimetric deviations from the optimal implant plan. A method for seed reconstruction from segmented C-arm fluoroscopy images is proposed. The 3D coordinates of the implanted seeds can be calculated upon resolving the correspondence of seeds in multiple x-ray images. We formalize seed-matching as a combinatorial optimization problem, which has salient features: (a) extensively studied solutions by the computer science community; (b) proof for the nonexistence of any polynomial time exact algorithm; and (c) a practical pseudo-polynomial algorithm that mostly runs in O(N3) time using any number of images. We prove that two images are insufficient to correctly match the seeds, while a third image renders the matching problem to be of nonpolynomial complexity. We utilize the special structure of the problem and propose a pseudopolynomial time algorithm. Using three presegmented images, matching and reconstruction of brachytherapy seeds using the Hungarian algorithm achieved complete matching in simulation experiments; and 98.5% in phantom experiments. 3D reconstruction error for correctly matched seeds has a mean of 0.63 mm, and 0.9 mm for incorrectly matched seeds. The maximum seed reconstruction error in each implant was typically around 1.32 mm. Both on synthetic data and in phantom experiments, matching rate and reconstruction error achieved using presegmented images was found to be sufficient for prostate brachytherapy. The algorithm is extendable to deal with arbitrary number of images without any loss in speed or accuracy. The algorithm is sufficiently generic to provide a practical solution to any correspondence problem, across different imaging modalities and features.

KW - C-arm

KW - Fluoroscopy

KW - Prostate brachytherapy

KW - Radiation planning

KW - Reconstruction

KW - Seed matching

UR - http://www.scopus.com/inward/record.url?scp=28044456093&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=28044456093&partnerID=8YFLogxK

U2 - 10.1118/1.2104087

DO - 10.1118/1.2104087

M3 - Article

C2 - 16372418

AN - SCOPUS:28044456093

VL - 32

SP - 3475

EP - 3492

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 11

ER -