Accurate Segmentation of CT Male Pelvic Organs via Regression-Based Deformable Models and Multi-Task Random Forests

Yaozong Gao, Yeqin Shao, Jun Lian, Andrew Z. Wang, Ronald C. Chen, Dinggang Shen

Research output: Contribution to journalArticle

33 Citations (Scopus)

Abstract

Segmenting male pelvic organs from CT images is a prerequisite for prostate cancer radiotherapy. The efficacy of radiation treatment highly depends on segmentation accuracy. However, accurate segmentation of male pelvic organs is challenging due to low tissue contrast of CT images, as well as large variations of shape and appearance of the pelvic organs. Among existing segmentation methods, deformable models are the most popular, as shape prior can be easily incorporated to regularize the segmentation. Nonetheless, the sensitivity to initialization often limits their performance, especially for segmenting organs with large shape variations. In this paper, we propose a novel approach to guide deformable models, thus making them robust against arbitrary initializations. Specifically, we learn a displacement regressor, which predicts 3D displacement from any image voxel to the target organ boundary based on the local patch appearance. This regressor provides a non-local external force for each vertex of deformable model, thus overcoming the initialization problem suffered by the traditional deformable models. To learn a reliable displacement regressor, two strategies are particularly proposed. 1) A multi-task random forest is proposed to learn the displacement regressor jointly with the organ classifier; 2) an auto-context model is used to iteratively enforce structural information during voxel-wise prediction. Extensive experiments on 313 planning CT scans of 313 patients show that our method achieves better results than alternative classification or regression based methods, and also several other existing methods in CT pelvic organ segmentation.

Original languageEnglish
Article number7384757
Pages (from-to)1532-1543
Number of pages12
JournalIEEE Transactions on Medical Imaging
Volume35
Issue number6
DOIs
Publication statusPublished - 2016 Jun 1

Fingerprint

Computerized tomography
Radiotherapy
Prostatic Neoplasms
Radiation
Classifiers
Forests
Tissue
Planning
Experiments

Keywords

  • computed tomography
  • deformable model
  • Image segmentation
  • machine learning
  • pelvic organs
  • prostate cancer
  • random forest

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Computer Science Applications
  • Radiological and Ultrasound Technology
  • Software

Cite this

Accurate Segmentation of CT Male Pelvic Organs via Regression-Based Deformable Models and Multi-Task Random Forests. / Gao, Yaozong; Shao, Yeqin; Lian, Jun; Wang, Andrew Z.; Chen, Ronald C.; Shen, Dinggang.

In: IEEE Transactions on Medical Imaging, Vol. 35, No. 6, 7384757, 01.06.2016, p. 1532-1543.

Research output: Contribution to journalArticle

Gao, Yaozong ; Shao, Yeqin ; Lian, Jun ; Wang, Andrew Z. ; Chen, Ronald C. ; Shen, Dinggang. / Accurate Segmentation of CT Male Pelvic Organs via Regression-Based Deformable Models and Multi-Task Random Forests. In: IEEE Transactions on Medical Imaging. 2016 ; Vol. 35, No. 6. pp. 1532-1543.
@article{72a97402caa64e93937b0b019d9b8bf3,
title = "Accurate Segmentation of CT Male Pelvic Organs via Regression-Based Deformable Models and Multi-Task Random Forests",
abstract = "Segmenting male pelvic organs from CT images is a prerequisite for prostate cancer radiotherapy. The efficacy of radiation treatment highly depends on segmentation accuracy. However, accurate segmentation of male pelvic organs is challenging due to low tissue contrast of CT images, as well as large variations of shape and appearance of the pelvic organs. Among existing segmentation methods, deformable models are the most popular, as shape prior can be easily incorporated to regularize the segmentation. Nonetheless, the sensitivity to initialization often limits their performance, especially for segmenting organs with large shape variations. In this paper, we propose a novel approach to guide deformable models, thus making them robust against arbitrary initializations. Specifically, we learn a displacement regressor, which predicts 3D displacement from any image voxel to the target organ boundary based on the local patch appearance. This regressor provides a non-local external force for each vertex of deformable model, thus overcoming the initialization problem suffered by the traditional deformable models. To learn a reliable displacement regressor, two strategies are particularly proposed. 1) A multi-task random forest is proposed to learn the displacement regressor jointly with the organ classifier; 2) an auto-context model is used to iteratively enforce structural information during voxel-wise prediction. Extensive experiments on 313 planning CT scans of 313 patients show that our method achieves better results than alternative classification or regression based methods, and also several other existing methods in CT pelvic organ segmentation.",
keywords = "computed tomography, deformable model, Image segmentation, machine learning, pelvic organs, prostate cancer, random forest",
author = "Yaozong Gao and Yeqin Shao and Jun Lian and Wang, {Andrew Z.} and Chen, {Ronald C.} and Dinggang Shen",
year = "2016",
month = "6",
day = "1",
doi = "10.1109/TMI.2016.2519264",
language = "English",
volume = "35",
pages = "1532--1543",
journal = "IEEE Transactions on Medical Imaging",
issn = "0278-0062",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "6",

}

TY - JOUR

T1 - Accurate Segmentation of CT Male Pelvic Organs via Regression-Based Deformable Models and Multi-Task Random Forests

AU - Gao, Yaozong

AU - Shao, Yeqin

AU - Lian, Jun

AU - Wang, Andrew Z.

AU - Chen, Ronald C.

AU - Shen, Dinggang

PY - 2016/6/1

Y1 - 2016/6/1

N2 - Segmenting male pelvic organs from CT images is a prerequisite for prostate cancer radiotherapy. The efficacy of radiation treatment highly depends on segmentation accuracy. However, accurate segmentation of male pelvic organs is challenging due to low tissue contrast of CT images, as well as large variations of shape and appearance of the pelvic organs. Among existing segmentation methods, deformable models are the most popular, as shape prior can be easily incorporated to regularize the segmentation. Nonetheless, the sensitivity to initialization often limits their performance, especially for segmenting organs with large shape variations. In this paper, we propose a novel approach to guide deformable models, thus making them robust against arbitrary initializations. Specifically, we learn a displacement regressor, which predicts 3D displacement from any image voxel to the target organ boundary based on the local patch appearance. This regressor provides a non-local external force for each vertex of deformable model, thus overcoming the initialization problem suffered by the traditional deformable models. To learn a reliable displacement regressor, two strategies are particularly proposed. 1) A multi-task random forest is proposed to learn the displacement regressor jointly with the organ classifier; 2) an auto-context model is used to iteratively enforce structural information during voxel-wise prediction. Extensive experiments on 313 planning CT scans of 313 patients show that our method achieves better results than alternative classification or regression based methods, and also several other existing methods in CT pelvic organ segmentation.

AB - Segmenting male pelvic organs from CT images is a prerequisite for prostate cancer radiotherapy. The efficacy of radiation treatment highly depends on segmentation accuracy. However, accurate segmentation of male pelvic organs is challenging due to low tissue contrast of CT images, as well as large variations of shape and appearance of the pelvic organs. Among existing segmentation methods, deformable models are the most popular, as shape prior can be easily incorporated to regularize the segmentation. Nonetheless, the sensitivity to initialization often limits their performance, especially for segmenting organs with large shape variations. In this paper, we propose a novel approach to guide deformable models, thus making them robust against arbitrary initializations. Specifically, we learn a displacement regressor, which predicts 3D displacement from any image voxel to the target organ boundary based on the local patch appearance. This regressor provides a non-local external force for each vertex of deformable model, thus overcoming the initialization problem suffered by the traditional deformable models. To learn a reliable displacement regressor, two strategies are particularly proposed. 1) A multi-task random forest is proposed to learn the displacement regressor jointly with the organ classifier; 2) an auto-context model is used to iteratively enforce structural information during voxel-wise prediction. Extensive experiments on 313 planning CT scans of 313 patients show that our method achieves better results than alternative classification or regression based methods, and also several other existing methods in CT pelvic organ segmentation.

KW - computed tomography

KW - deformable model

KW - Image segmentation

KW - machine learning

KW - pelvic organs

KW - prostate cancer

KW - random forest

UR - http://www.scopus.com/inward/record.url?scp=84974575225&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84974575225&partnerID=8YFLogxK

U2 - 10.1109/TMI.2016.2519264

DO - 10.1109/TMI.2016.2519264

M3 - Article

C2 - 26800531

AN - SCOPUS:84974575225

VL - 35

SP - 1532

EP - 1543

JO - IEEE Transactions on Medical Imaging

JF - IEEE Transactions on Medical Imaging

SN - 0278-0062

IS - 6

M1 - 7384757

ER -