Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks

Sihang Zhou, Dong Nie, Ehsan Adeli, Yaozong Gao, Li Wang, Jianping Yin, Dinggang Shen

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

2 Citations (Scopus)

Abstract

Image segmentation is a crucial step in many computer-aided medical image analysis tasks, e.g., automated radiation therapy. However, low tissue-contrast and large amounts of artifacts in medical images, i.e., CT or MR images, corrupt the true boundaries of the target tissues and adversely influence the precision of boundary localization in segmentation. To precisely locate blurry and missing boundaries, human observers often use high-resolution context information from neighboring regions. To extract such information and achieve fine-grained segmentation (high accuracy on the boundary regions and small-scale targets), we propose a novel hierarchical dilated network. In the hierarchy, to maintain precise location information, we adopt dilated residual convolutional blocks as basic building blocks to reduce the dependency of the network on downsampling for receptive field enlargement and semantic information extraction. Then, by concatenating the intermediate feature maps of the serially-connected dilated residual convolutional blocks, the resultant hierarchical dilated module (HD-module) can encourage more smooth information flow and better utilization of both high-level semantic information and low-level textural information. Finally, we integrate several HD-modules in different resolutions in a parallel connection fashion to finely collect information from multiple (more than 12) scales for the network. The integration is defined by a novel late fusion module proposed in this paper. Experimental results on pelvic organ CT image segmentation demonstrate the superior performance of our proposed algorithm to the state-of-the-art deep learning segmentation algorithms, especially in localizing the organ boundaries.

Original languageEnglish
Title of host publicationMedical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings
EditorsAlejandro F. Frangi, Gabor Fichtinger, Julia A. Schnabel, Carlos Alberola-López, Christos Davatzikos
PublisherSpringer Verlag
Pages488-496
Number of pages9
ISBN (Print)9783030009366
DOIs
Publication statusPublished - 2018 Jan 1
Externally publishedYes
Event21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018 - Granada, Spain
Duration: 2018 Sep 162018 Sep 20

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11073 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018
CountrySpain
CityGranada
Period18/9/1618/9/20

Fingerprint

Hierarchical Networks
Image segmentation
Segmentation
Semantics
Neural Networks
Tissue
Neural networks
Radiotherapy
Image analysis
Fusion reactions
Image Segmentation
Module
Medical Image Analysis
Radiation Therapy
Receptive Field
Target
CT Image
Enlargement
Information Extraction
Information Flow

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Zhou, S., Nie, D., Adeli, E., Gao, Y., Wang, L., Yin, J., & Shen, D. (2018). Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks. In A. F. Frangi, G. Fichtinger, J. A. Schnabel, C. Alberola-López, & C. Davatzikos (Eds.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings (pp. 488-496). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11073 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-00937-3_56

Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks. / Zhou, Sihang; Nie, Dong; Adeli, Ehsan; Gao, Yaozong; Wang, Li; Yin, Jianping; Shen, Dinggang.

Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. ed. / Alejandro F. Frangi; Gabor Fichtinger; Julia A. Schnabel; Carlos Alberola-López; Christos Davatzikos. Springer Verlag, 2018. p. 488-496 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11073 LNCS).

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

Zhou, S, Nie, D, Adeli, E, Gao, Y, Wang, L, Yin, J & Shen, D 2018, Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks. in AF Frangi, G Fichtinger, JA Schnabel, C Alberola-López & C Davatzikos (eds), Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11073 LNCS, Springer Verlag, pp. 488-496, 21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018, Granada, Spain, 18/9/16. https://doi.org/10.1007/978-3-030-00937-3_56
Zhou S, Nie D, Adeli E, Gao Y, Wang L, Yin J et al. Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks. In Frangi AF, Fichtinger G, Schnabel JA, Alberola-López C, Davatzikos C, editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. Springer Verlag. 2018. p. 488-496. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-00937-3_56
Zhou, Sihang ; Nie, Dong ; Adeli, Ehsan ; Gao, Yaozong ; Wang, Li ; Yin, Jianping ; Shen, Dinggang. / Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks. Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. editor / Alejandro F. Frangi ; Gabor Fichtinger ; Julia A. Schnabel ; Carlos Alberola-López ; Christos Davatzikos. Springer Verlag, 2018. pp. 488-496 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{daccda4fbc17422ba5a2816ff2c46137,
title = "Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks",
abstract = "Image segmentation is a crucial step in many computer-aided medical image analysis tasks, e.g., automated radiation therapy. However, low tissue-contrast and large amounts of artifacts in medical images, i.e., CT or MR images, corrupt the true boundaries of the target tissues and adversely influence the precision of boundary localization in segmentation. To precisely locate blurry and missing boundaries, human observers often use high-resolution context information from neighboring regions. To extract such information and achieve fine-grained segmentation (high accuracy on the boundary regions and small-scale targets), we propose a novel hierarchical dilated network. In the hierarchy, to maintain precise location information, we adopt dilated residual convolutional blocks as basic building blocks to reduce the dependency of the network on downsampling for receptive field enlargement and semantic information extraction. Then, by concatenating the intermediate feature maps of the serially-connected dilated residual convolutional blocks, the resultant hierarchical dilated module (HD-module) can encourage more smooth information flow and better utilization of both high-level semantic information and low-level textural information. Finally, we integrate several HD-modules in different resolutions in a parallel connection fashion to finely collect information from multiple (more than 12) scales for the network. The integration is defined by a novel late fusion module proposed in this paper. Experimental results on pelvic organ CT image segmentation demonstrate the superior performance of our proposed algorithm to the state-of-the-art deep learning segmentation algorithms, especially in localizing the organ boundaries.",
author = "Sihang Zhou and Dong Nie and Ehsan Adeli and Yaozong Gao and Li Wang and Jianping Yin and Dinggang Shen",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-00937-3_56",
language = "English",
isbn = "9783030009366",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "488--496",
editor = "Frangi, {Alejandro F.} and Gabor Fichtinger and Schnabel, {Julia A.} and Carlos Alberola-L{\'o}pez and Christos Davatzikos",
booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings",

}

TY - GEN

T1 - Fine-Grained Segmentation Using Hierarchical Dilated Neural Networks

AU - Zhou, Sihang

AU - Nie, Dong

AU - Adeli, Ehsan

AU - Gao, Yaozong

AU - Wang, Li

AU - Yin, Jianping

AU - Shen, Dinggang

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Image segmentation is a crucial step in many computer-aided medical image analysis tasks, e.g., automated radiation therapy. However, low tissue-contrast and large amounts of artifacts in medical images, i.e., CT or MR images, corrupt the true boundaries of the target tissues and adversely influence the precision of boundary localization in segmentation. To precisely locate blurry and missing boundaries, human observers often use high-resolution context information from neighboring regions. To extract such information and achieve fine-grained segmentation (high accuracy on the boundary regions and small-scale targets), we propose a novel hierarchical dilated network. In the hierarchy, to maintain precise location information, we adopt dilated residual convolutional blocks as basic building blocks to reduce the dependency of the network on downsampling for receptive field enlargement and semantic information extraction. Then, by concatenating the intermediate feature maps of the serially-connected dilated residual convolutional blocks, the resultant hierarchical dilated module (HD-module) can encourage more smooth information flow and better utilization of both high-level semantic information and low-level textural information. Finally, we integrate several HD-modules in different resolutions in a parallel connection fashion to finely collect information from multiple (more than 12) scales for the network. The integration is defined by a novel late fusion module proposed in this paper. Experimental results on pelvic organ CT image segmentation demonstrate the superior performance of our proposed algorithm to the state-of-the-art deep learning segmentation algorithms, especially in localizing the organ boundaries.

AB - Image segmentation is a crucial step in many computer-aided medical image analysis tasks, e.g., automated radiation therapy. However, low tissue-contrast and large amounts of artifacts in medical images, i.e., CT or MR images, corrupt the true boundaries of the target tissues and adversely influence the precision of boundary localization in segmentation. To precisely locate blurry and missing boundaries, human observers often use high-resolution context information from neighboring regions. To extract such information and achieve fine-grained segmentation (high accuracy on the boundary regions and small-scale targets), we propose a novel hierarchical dilated network. In the hierarchy, to maintain precise location information, we adopt dilated residual convolutional blocks as basic building blocks to reduce the dependency of the network on downsampling for receptive field enlargement and semantic information extraction. Then, by concatenating the intermediate feature maps of the serially-connected dilated residual convolutional blocks, the resultant hierarchical dilated module (HD-module) can encourage more smooth information flow and better utilization of both high-level semantic information and low-level textural information. Finally, we integrate several HD-modules in different resolutions in a parallel connection fashion to finely collect information from multiple (more than 12) scales for the network. The integration is defined by a novel late fusion module proposed in this paper. Experimental results on pelvic organ CT image segmentation demonstrate the superior performance of our proposed algorithm to the state-of-the-art deep learning segmentation algorithms, especially in localizing the organ boundaries.

UR - http://www.scopus.com/inward/record.url?scp=85053850188&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053850188&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-00937-3_56

DO - 10.1007/978-3-030-00937-3_56

M3 - Conference contribution

AN - SCOPUS:85053850188

SN - 9783030009366

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

SP - 488

EP - 496

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings

A2 - Frangi, Alejandro F.

A2 - Fichtinger, Gabor

A2 - Schnabel, Julia A.

A2 - Alberola-López, Carlos

A2 - Davatzikos, Christos

PB - Springer Verlag

ER -