Hierarchical shape statistical model for segmentation of lung fields in chest radiographs

Yonghong Shi, Dinggang Shen

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

19 Citations (Scopus)

Abstract

The standard Active Shape Model (ASM) generally uses a whole population to train a single PCA-based shape model for segmentation of all testing samples. Since some testing samples can be similar to only sub-population of training samples, it will be more effective if particular shape statistics extracted from the respective sub-population can be used for guiding image segmentation. Accordingly, we design a set of hierarchical shape statistical models, including a whole-population shape model and a series of sub-population models. The whole-population shape model is used to guide the initial segmentation of the testing sample, and the initial segmentation result is then used to select a suitable sub-population shape model according to the shape similarity between the testing sample and each sub-population. By using the selected sub-population shape model, the segmentation result can be further refined. To achieve this segmentation process, several particular steps are designed next. First, all linearly aligned samples in the whole population are used to generate a whole-population shape model. Second, an affinity propagation method is used to cluster all linearly aligned samples into several clusters, to determine the samples belonging to the same sub-populations. Third, the original samples of each sub-population are linearly aligned to their own mean shape, and the respective sub-population shape model is built using the newly aligned samples in this sub-population. By using all these three steps, we can generate hierarchical shape statistical models to guide image segmentation. Experimental results show that the proposed method can significantly improve the segmentation performance, compared to conventional ASM.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages417-424
Number of pages8
Volume5241 LNCS
EditionPART 1
DOIs
Publication statusPublished - 2008 Dec 1
Externally publishedYes
Event11th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2008 - New York, NY, United States
Duration: 2008 Sep 62008 Sep 10

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume5241 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other11th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2008
CountryUnited States
CityNew York, NY
Period08/9/608/9/10

Fingerprint

Lung
Statistical Model
Segmentation
Testing
Image segmentation
Active Shape Model
Linearly
Statistical Models
Image Segmentation
Model
Training Samples
Population Model
Statistics
Affine transformation
Standard Model

Keywords

  • Active shape model
  • Chest radiograph
  • Hierarchical shape statistics

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Shi, Y., & Shen, D. (2008). Hierarchical shape statistical model for segmentation of lung fields in chest radiographs. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 1 ed., Vol. 5241 LNCS, pp. 417-424). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5241 LNCS, No. PART 1). https://doi.org/10.1007/978-3-540-85988-8_50

Hierarchical shape statistical model for segmentation of lung fields in chest radiographs. / Shi, Yonghong; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5241 LNCS PART 1. ed. 2008. p. 417-424 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5241 LNCS, No. PART 1).

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

Shi, Y & Shen, D 2008, Hierarchical shape statistical model for segmentation of lung fields in chest radiographs. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 edn, vol. 5241 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 5241 LNCS, pp. 417-424, 11th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2008, New York, NY, United States, 08/9/6. https://doi.org/10.1007/978-3-540-85988-8_50
Shi Y, Shen D. Hierarchical shape statistical model for segmentation of lung fields in chest radiographs. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 ed. Vol. 5241 LNCS. 2008. p. 417-424. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-540-85988-8_50
Shi, Yonghong ; Shen, Dinggang. / Hierarchical shape statistical model for segmentation of lung fields in chest radiographs. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5241 LNCS PART 1. ed. 2008. pp. 417-424 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{58bdf7d9abb04c13bab8e5f838593be4,
title = "Hierarchical shape statistical model for segmentation of lung fields in chest radiographs",
abstract = "The standard Active Shape Model (ASM) generally uses a whole population to train a single PCA-based shape model for segmentation of all testing samples. Since some testing samples can be similar to only sub-population of training samples, it will be more effective if particular shape statistics extracted from the respective sub-population can be used for guiding image segmentation. Accordingly, we design a set of hierarchical shape statistical models, including a whole-population shape model and a series of sub-population models. The whole-population shape model is used to guide the initial segmentation of the testing sample, and the initial segmentation result is then used to select a suitable sub-population shape model according to the shape similarity between the testing sample and each sub-population. By using the selected sub-population shape model, the segmentation result can be further refined. To achieve this segmentation process, several particular steps are designed next. First, all linearly aligned samples in the whole population are used to generate a whole-population shape model. Second, an affinity propagation method is used to cluster all linearly aligned samples into several clusters, to determine the samples belonging to the same sub-populations. Third, the original samples of each sub-population are linearly aligned to their own mean shape, and the respective sub-population shape model is built using the newly aligned samples in this sub-population. By using all these three steps, we can generate hierarchical shape statistical models to guide image segmentation. Experimental results show that the proposed method can significantly improve the segmentation performance, compared to conventional ASM.",
keywords = "Active shape model, Chest radiograph, Hierarchical shape statistics",
author = "Yonghong Shi and Dinggang Shen",
year = "2008",
month = "12",
day = "1",
doi = "10.1007/978-3-540-85988-8_50",
language = "English",
isbn = "354085987X",
volume = "5241 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "417--424",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 1",

}

TY - GEN

T1 - Hierarchical shape statistical model for segmentation of lung fields in chest radiographs

AU - Shi, Yonghong

AU - Shen, Dinggang

PY - 2008/12/1

Y1 - 2008/12/1

N2 - The standard Active Shape Model (ASM) generally uses a whole population to train a single PCA-based shape model for segmentation of all testing samples. Since some testing samples can be similar to only sub-population of training samples, it will be more effective if particular shape statistics extracted from the respective sub-population can be used for guiding image segmentation. Accordingly, we design a set of hierarchical shape statistical models, including a whole-population shape model and a series of sub-population models. The whole-population shape model is used to guide the initial segmentation of the testing sample, and the initial segmentation result is then used to select a suitable sub-population shape model according to the shape similarity between the testing sample and each sub-population. By using the selected sub-population shape model, the segmentation result can be further refined. To achieve this segmentation process, several particular steps are designed next. First, all linearly aligned samples in the whole population are used to generate a whole-population shape model. Second, an affinity propagation method is used to cluster all linearly aligned samples into several clusters, to determine the samples belonging to the same sub-populations. Third, the original samples of each sub-population are linearly aligned to their own mean shape, and the respective sub-population shape model is built using the newly aligned samples in this sub-population. By using all these three steps, we can generate hierarchical shape statistical models to guide image segmentation. Experimental results show that the proposed method can significantly improve the segmentation performance, compared to conventional ASM.

AB - The standard Active Shape Model (ASM) generally uses a whole population to train a single PCA-based shape model for segmentation of all testing samples. Since some testing samples can be similar to only sub-population of training samples, it will be more effective if particular shape statistics extracted from the respective sub-population can be used for guiding image segmentation. Accordingly, we design a set of hierarchical shape statistical models, including a whole-population shape model and a series of sub-population models. The whole-population shape model is used to guide the initial segmentation of the testing sample, and the initial segmentation result is then used to select a suitable sub-population shape model according to the shape similarity between the testing sample and each sub-population. By using the selected sub-population shape model, the segmentation result can be further refined. To achieve this segmentation process, several particular steps are designed next. First, all linearly aligned samples in the whole population are used to generate a whole-population shape model. Second, an affinity propagation method is used to cluster all linearly aligned samples into several clusters, to determine the samples belonging to the same sub-populations. Third, the original samples of each sub-population are linearly aligned to their own mean shape, and the respective sub-population shape model is built using the newly aligned samples in this sub-population. By using all these three steps, we can generate hierarchical shape statistical models to guide image segmentation. Experimental results show that the proposed method can significantly improve the segmentation performance, compared to conventional ASM.

KW - Active shape model

KW - Chest radiograph

KW - Hierarchical shape statistics

UR - http://www.scopus.com/inward/record.url?scp=58849128362&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=58849128362&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-85988-8_50

DO - 10.1007/978-3-540-85988-8_50

M3 - Conference contribution

C2 - 18979774

AN - SCOPUS:58849128362

SN - 354085987X

SN - 9783540859871

VL - 5241 LNCS

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

SP - 417

EP - 424

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

ER -