Automatic liver vessel segmentation using 3D region growing and hybrid active contour model

Ye-Zhan Zeng; Sheng-Hui Liao; Ping Tang; Yu-Qian Zhao; Miao Liao; Yan Chen; Yi-Xiong Liang

doi:10.1016/j.compbiomed.2018.04.014

Automatic liver vessel segmentation using 3D region growing and hybrid active contour model

Comput Biol Med. 2018 Jun 1:97:63-73. doi: 10.1016/j.compbiomed.2018.04.014. Epub 2018 Apr 24.

Authors

Ye-Zhan Zeng¹, Sheng-Hui Liao², Ping Tang¹, Yu-Qian Zhao³, Miao Liao⁴, Yan Chen⁵, Yi-Xiong Liang⁶

Affiliations

¹ School of Information Science and Engineering, Central South University, Changsha, 410083, China; Department of Biomedical Engineering, Central South University, Changsha, 410083, China.
² School of Information Science and Engineering, Central South University, Changsha, 410083, China. Electronic address: [email protected].
³ School of Information Science and Engineering, Central South University, Changsha, 410083, China; Department of Biomedical Engineering, Central South University, Changsha, 410083, China. Electronic address: [email protected].
⁴ School of Computer Science and Engineering, Hunan University of Science and Technology, Xiangtan, 411201, China.
⁵ Applied Vision Research Centre, Loughborough University, Loughborough, UK.
⁶ School of Information Science and Engineering, Central South University, Changsha, 410083, China.

PMID: 29709715
DOI: 10.1016/j.compbiomed.2018.04.014

Abstract

This paper proposes a new automatic method for liver vessel segmentation by exploiting intensity and shape constraints of 3D vessels. The core of the proposed method is to apply two different strategies: 3D region growing facilitated by bi-Gaussian filter for thin vessel segmentation, and hybrid active contour model combined with K-means clustering for thick vessel segmentation. They are then integrated to generate final segmentation results. The proposed method is validated on abdominal computed tomography angiography (CTA) images, and obtains an average accuracy, sensitivity, specificity, Dice, Jaccard, and RMSD of 98.2%, 68.3%, 99.2%, 73.0%, 66.1%, and 2.56 mm, respectively. Experimental results show that our method is capable of segmenting complex liver vessels with more continuous and complete thin vessel details, and outperforms several existing 3D vessel segmentation algorithms.

Keywords: 3D region growing; Bi-Gaussian filter; Hybrid active contour model; Liver vessel segmentation.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Algorithms
Computed Tomography Angiography / methods*
Hepatic Artery / diagnostic imaging*
Hepatic Veins / diagnostic imaging*
Humans
Imaging, Three-Dimensional / methods*
Liver* / blood supply
Liver* / diagnostic imaging