Dense deformation reconstruction via sparse coding

Yonghong Shi, Guorong Wu, Zhijian Song, Dinggang Shen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many image registration algorithms need to interpolate dense deformations from a small set of sparse deformations or correspondences established on the landmark points. Previous methods generally use a certain pre-defined deformation model, e.g., B-Spline or Thin-Plate Spline, for dense deformation interpolation, which may affect the final registration accuracy since the actual deformation may not exactly follow the pre-defined model. To address this issue, we propose a novel leaning-based method to represent the to-be-estimated dense deformations as a linear combination of sample dense deformations in the pre-constructed dictionary, with the combination coefficients computed from sparse representation of their respective correspondences on the same set of landmarks. Specifically, in the training stage, for each training image, we register it to the selected template by a certain registration method and obtain correspondences on a fixed set of landmarks in the template, as well as the respective dense deformation field. Then, we can build two dictionaries to, respectively, save the landmark correspondences and their dense deformations from all training images at the same indexing order. Thus, in the application stage, after estimating the landmark correspondences for a new subject, we can first represent them by all instances in the dictionary of landmark correspondences. Then, the estimated sparse coefficients can be used to reconstruct the dense deformation field of the new subject by fusing the corresponding instances in the dictionary of dense deformations. We have demonstrated the advantage of our proposed deformation interpolation method in two applications, i.e., CT prostate registration in the radiotherapy and MR brain registration in the neuroscience study. In both applications, our learning-based method can achieve higher accuracy and potentially faster computation, compared to the conventional method.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages36-44
Number of pages9
Volume7588 LNCS
DOIs
Publication statusPublished - 2012 Nov 30
Externally publishedYes
Event3rd International Workshop on Machine Learning in Medical Imaging, MLMI 2012, Held in conjunction with the 15th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2012 - Nice, France
Duration: 2012 Oct 12012 Oct 1

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7588 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other3rd International Workshop on Machine Learning in Medical Imaging, MLMI 2012, Held in conjunction with the 15th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2012
CountryFrance
CityNice
Period12/10/112/10/1

Fingerprint

Sparse Coding
Landmarks
Correspondence
Glossaries
Registration
Splines
Template
Interpolation
Interpolate
Thin-plate Spline
Radiotherapy
Neuroscience
Sparse Representation
Image registration
Interpolation Method
Image Registration
Coefficient
B-spline
Indexing
Linear Combination

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Shi, Y., Wu, G., Song, Z., & Shen, D. (2012). Dense deformation reconstruction via sparse coding. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7588 LNCS, pp. 36-44). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7588 LNCS). https://doi.org/10.1007/978-3-642-35428-1_5

Dense deformation reconstruction via sparse coding. / Shi, Yonghong; Wu, Guorong; Song, Zhijian; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7588 LNCS 2012. p. 36-44 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7588 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Shi, Y, Wu, G, Song, Z & Shen, D 2012, Dense deformation reconstruction via sparse coding. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 7588 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7588 LNCS, pp. 36-44, 3rd International Workshop on Machine Learning in Medical Imaging, MLMI 2012, Held in conjunction with the 15th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2012, Nice, France, 12/10/1. https://doi.org/10.1007/978-3-642-35428-1_5
Shi Y, Wu G, Song Z, Shen D. Dense deformation reconstruction via sparse coding. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7588 LNCS. 2012. p. 36-44. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-35428-1_5
Shi, Yonghong ; Wu, Guorong ; Song, Zhijian ; Shen, Dinggang. / Dense deformation reconstruction via sparse coding. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7588 LNCS 2012. pp. 36-44 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{e30b047d888e4fbe9e642640aede6f14,
title = "Dense deformation reconstruction via sparse coding",
abstract = "Many image registration algorithms need to interpolate dense deformations from a small set of sparse deformations or correspondences established on the landmark points. Previous methods generally use a certain pre-defined deformation model, e.g., B-Spline or Thin-Plate Spline, for dense deformation interpolation, which may affect the final registration accuracy since the actual deformation may not exactly follow the pre-defined model. To address this issue, we propose a novel leaning-based method to represent the to-be-estimated dense deformations as a linear combination of sample dense deformations in the pre-constructed dictionary, with the combination coefficients computed from sparse representation of their respective correspondences on the same set of landmarks. Specifically, in the training stage, for each training image, we register it to the selected template by a certain registration method and obtain correspondences on a fixed set of landmarks in the template, as well as the respective dense deformation field. Then, we can build two dictionaries to, respectively, save the landmark correspondences and their dense deformations from all training images at the same indexing order. Thus, in the application stage, after estimating the landmark correspondences for a new subject, we can first represent them by all instances in the dictionary of landmark correspondences. Then, the estimated sparse coefficients can be used to reconstruct the dense deformation field of the new subject by fusing the corresponding instances in the dictionary of dense deformations. We have demonstrated the advantage of our proposed deformation interpolation method in two applications, i.e., CT prostate registration in the radiotherapy and MR brain registration in the neuroscience study. In both applications, our learning-based method can achieve higher accuracy and potentially faster computation, compared to the conventional method.",
author = "Yonghong Shi and Guorong Wu and Zhijian Song and Dinggang Shen",
year = "2012",
month = "11",
day = "30",
doi = "10.1007/978-3-642-35428-1_5",
language = "English",
isbn = "9783642354274",
volume = "7588 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "36--44",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Dense deformation reconstruction via sparse coding

AU - Shi, Yonghong

AU - Wu, Guorong

AU - Song, Zhijian

AU - Shen, Dinggang

PY - 2012/11/30

Y1 - 2012/11/30

N2 - Many image registration algorithms need to interpolate dense deformations from a small set of sparse deformations or correspondences established on the landmark points. Previous methods generally use a certain pre-defined deformation model, e.g., B-Spline or Thin-Plate Spline, for dense deformation interpolation, which may affect the final registration accuracy since the actual deformation may not exactly follow the pre-defined model. To address this issue, we propose a novel leaning-based method to represent the to-be-estimated dense deformations as a linear combination of sample dense deformations in the pre-constructed dictionary, with the combination coefficients computed from sparse representation of their respective correspondences on the same set of landmarks. Specifically, in the training stage, for each training image, we register it to the selected template by a certain registration method and obtain correspondences on a fixed set of landmarks in the template, as well as the respective dense deformation field. Then, we can build two dictionaries to, respectively, save the landmark correspondences and their dense deformations from all training images at the same indexing order. Thus, in the application stage, after estimating the landmark correspondences for a new subject, we can first represent them by all instances in the dictionary of landmark correspondences. Then, the estimated sparse coefficients can be used to reconstruct the dense deformation field of the new subject by fusing the corresponding instances in the dictionary of dense deformations. We have demonstrated the advantage of our proposed deformation interpolation method in two applications, i.e., CT prostate registration in the radiotherapy and MR brain registration in the neuroscience study. In both applications, our learning-based method can achieve higher accuracy and potentially faster computation, compared to the conventional method.

AB - Many image registration algorithms need to interpolate dense deformations from a small set of sparse deformations or correspondences established on the landmark points. Previous methods generally use a certain pre-defined deformation model, e.g., B-Spline or Thin-Plate Spline, for dense deformation interpolation, which may affect the final registration accuracy since the actual deformation may not exactly follow the pre-defined model. To address this issue, we propose a novel leaning-based method to represent the to-be-estimated dense deformations as a linear combination of sample dense deformations in the pre-constructed dictionary, with the combination coefficients computed from sparse representation of their respective correspondences on the same set of landmarks. Specifically, in the training stage, for each training image, we register it to the selected template by a certain registration method and obtain correspondences on a fixed set of landmarks in the template, as well as the respective dense deformation field. Then, we can build two dictionaries to, respectively, save the landmark correspondences and their dense deformations from all training images at the same indexing order. Thus, in the application stage, after estimating the landmark correspondences for a new subject, we can first represent them by all instances in the dictionary of landmark correspondences. Then, the estimated sparse coefficients can be used to reconstruct the dense deformation field of the new subject by fusing the corresponding instances in the dictionary of dense deformations. We have demonstrated the advantage of our proposed deformation interpolation method in two applications, i.e., CT prostate registration in the radiotherapy and MR brain registration in the neuroscience study. In both applications, our learning-based method can achieve higher accuracy and potentially faster computation, compared to the conventional method.

UR - http://www.scopus.com/inward/record.url?scp=84870019130&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84870019130&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-35428-1_5

DO - 10.1007/978-3-642-35428-1_5

M3 - Conference contribution

SN - 9783642354274

VL - 7588 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 36

EP - 44

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -