Multi-atlas context forests for knee MR image segmentation

Qin Liu, Qian Wang, Lichi Zhang, Yaozong Gao, Dinggang Shen

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

2 Citations (Scopus)

Abstract

It is important, yet a challenging procedure, to segment bones and cartilages from knee MR images. In this paper, we propose multi-atlas context forests to first segment bones and then segment cartilages. Specifically, for both the bone and cartilage segmentations, we iteratively train sets of random forests, based on training atlas images, to classify the individual voxels. The random forests rely on (1) the appearance features directly computed from images and also (2) the context features associated with tentative segmentation results, generated by the previous layer of random forest in the iterative framework. To extract context features, multiple atlases (with expert segmentation) are first registered, with the tentative segmentation result of the subject under consideration. Then, the spatial priors of anatomical labels of registered atlases are computed and used to calculate context features of the subject. Note that these multi-atlas context features will be iteratively refined based on the (updated) tentative segmentation result of the subject. As better segmentation result leads to more accurate registration between multiple atlases and the subject, context features will become increasingly more useful for the training of subsequent random forests in the iterative framework. As validated by experiments on the SKI10 dataset, our proposed method can achieve high segmentation accuracy.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages186-193
Number of pages8
Volume9352
ISBN (Print)9783319248875
DOIs
Publication statusPublished - 2015
Externally publishedYes
Event6th International Workshop on Machine Learning in Medical Imaging, MLMI 2015 and Held in Conjunction with 18th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2015 - Munich, Germany
Duration: 2015 Oct 52015 Oct 5

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9352
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other6th International Workshop on Machine Learning in Medical Imaging, MLMI 2015 and Held in Conjunction with 18th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2015
CountryGermany
CityMunich
Period15/10/515/10/5

Fingerprint

Atlas
Cartilage
Image segmentation
Image Segmentation
Bone
Segmentation
Random Forest
Labels
Voxel
Context
Registration
Experiments
Classify
Calculate
Experiment

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Liu, Q., Wang, Q., Zhang, L., Gao, Y., & Shen, D. (2015). Multi-atlas context forests for knee MR image segmentation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9352, pp. 186-193). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9352). Springer Verlag. https://doi.org/10.1007/978-3-319-24888-2_23

Multi-atlas context forests for knee MR image segmentation. / Liu, Qin; Wang, Qian; Zhang, Lichi; Gao, Yaozong; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9352 Springer Verlag, 2015. p. 186-193 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9352).

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

Liu, Q, Wang, Q, Zhang, L, Gao, Y & Shen, D 2015, Multi-atlas context forests for knee MR image segmentation. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 9352, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9352, Springer Verlag, pp. 186-193, 6th International Workshop on Machine Learning in Medical Imaging, MLMI 2015 and Held in Conjunction with 18th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2015, Munich, Germany, 15/10/5. https://doi.org/10.1007/978-3-319-24888-2_23
Liu Q, Wang Q, Zhang L, Gao Y, Shen D. Multi-atlas context forests for knee MR image segmentation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9352. Springer Verlag. 2015. p. 186-193. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-24888-2_23
Liu, Qin ; Wang, Qian ; Zhang, Lichi ; Gao, Yaozong ; Shen, Dinggang. / Multi-atlas context forests for knee MR image segmentation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9352 Springer Verlag, 2015. pp. 186-193 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{488d8890475d4f15b745e756b2e0ff52,
title = "Multi-atlas context forests for knee MR image segmentation",
abstract = "It is important, yet a challenging procedure, to segment bones and cartilages from knee MR images. In this paper, we propose multi-atlas context forests to first segment bones and then segment cartilages. Specifically, for both the bone and cartilage segmentations, we iteratively train sets of random forests, based on training atlas images, to classify the individual voxels. The random forests rely on (1) the appearance features directly computed from images and also (2) the context features associated with tentative segmentation results, generated by the previous layer of random forest in the iterative framework. To extract context features, multiple atlases (with expert segmentation) are first registered, with the tentative segmentation result of the subject under consideration. Then, the spatial priors of anatomical labels of registered atlases are computed and used to calculate context features of the subject. Note that these multi-atlas context features will be iteratively refined based on the (updated) tentative segmentation result of the subject. As better segmentation result leads to more accurate registration between multiple atlases and the subject, context features will become increasingly more useful for the training of subsequent random forests in the iterative framework. As validated by experiments on the SKI10 dataset, our proposed method can achieve high segmentation accuracy.",
author = "Qin Liu and Qian Wang and Lichi Zhang and Yaozong Gao and Dinggang Shen",
year = "2015",
doi = "10.1007/978-3-319-24888-2_23",
language = "English",
isbn = "9783319248875",
volume = "9352",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "186--193",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Multi-atlas context forests for knee MR image segmentation

AU - Liu, Qin

AU - Wang, Qian

AU - Zhang, Lichi

AU - Gao, Yaozong

AU - Shen, Dinggang

PY - 2015

Y1 - 2015

N2 - It is important, yet a challenging procedure, to segment bones and cartilages from knee MR images. In this paper, we propose multi-atlas context forests to first segment bones and then segment cartilages. Specifically, for both the bone and cartilage segmentations, we iteratively train sets of random forests, based on training atlas images, to classify the individual voxels. The random forests rely on (1) the appearance features directly computed from images and also (2) the context features associated with tentative segmentation results, generated by the previous layer of random forest in the iterative framework. To extract context features, multiple atlases (with expert segmentation) are first registered, with the tentative segmentation result of the subject under consideration. Then, the spatial priors of anatomical labels of registered atlases are computed and used to calculate context features of the subject. Note that these multi-atlas context features will be iteratively refined based on the (updated) tentative segmentation result of the subject. As better segmentation result leads to more accurate registration between multiple atlases and the subject, context features will become increasingly more useful for the training of subsequent random forests in the iterative framework. As validated by experiments on the SKI10 dataset, our proposed method can achieve high segmentation accuracy.

AB - It is important, yet a challenging procedure, to segment bones and cartilages from knee MR images. In this paper, we propose multi-atlas context forests to first segment bones and then segment cartilages. Specifically, for both the bone and cartilage segmentations, we iteratively train sets of random forests, based on training atlas images, to classify the individual voxels. The random forests rely on (1) the appearance features directly computed from images and also (2) the context features associated with tentative segmentation results, generated by the previous layer of random forest in the iterative framework. To extract context features, multiple atlases (with expert segmentation) are first registered, with the tentative segmentation result of the subject under consideration. Then, the spatial priors of anatomical labels of registered atlases are computed and used to calculate context features of the subject. Note that these multi-atlas context features will be iteratively refined based on the (updated) tentative segmentation result of the subject. As better segmentation result leads to more accurate registration between multiple atlases and the subject, context features will become increasingly more useful for the training of subsequent random forests in the iterative framework. As validated by experiments on the SKI10 dataset, our proposed method can achieve high segmentation accuracy.

UR - http://www.scopus.com/inward/record.url?scp=84952060545&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84952060545&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-24888-2_23

DO - 10.1007/978-3-319-24888-2_23

M3 - Conference contribution

AN - SCOPUS:84952060545

SN - 9783319248875

VL - 9352

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

SP - 186

EP - 193

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

PB - Springer Verlag

ER -