Segmentation and classification of colon glands with deep convolutional neural networks and total variation regularization

Philipp Kainz, Michael Pfeiffer, Martin Urschler

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Segmentation of histopathology sections is a necessary preprocessing step for digital pathology. Due to the large variability of biological tissue, machine learning techniques have shown superior performance over conventional image processing methods. Here we present our deep neural network-based approach for segmentation and classification of glands in tissue of benign and malignant colorectal cancer, which was developed to participate in the GlaS@MICCAI2015 colon gland segmentation challenge. We use two distinct deep convolutional neural networks (CNN) for pixel-wise classification of Hematoxylin-Eosin stained images. While the first classifier separates glands from background, the second classifier identifies gland-separating structures. In a subsequent step, a figure-ground segmentation based on weighted total variation produces the final segmentation result by regularizing the CNN predictions. We present both quantitative and qualitative segmentation results on the recently released and publicly available Warwick-QU colon adenocarcinoma dataset associated with the GlaS@MICCAI2015 challenge and compare our approach to the simultaneously developed other approaches that participated in the same challenge. On two test sets, we demonstrate our segmentation performance and show that we achieve a tissue classification accuracy of 98% and 95%, making use of the inherent capability of our system to distinguish between benign and malignant tissue. Our results show that deep learning approaches can yield highly accurate and reproducible results for biomedical image analysis, with the potential to significantly improve the quality and speed of medical diagnoses.

LanguageEnglish
Article numbere3874
Number of pages28
JournalPeerJ
Volume5
DOIs
StatusPublished - 3 Oct 2017

Fingerprint

neural networks
colon
Colon
Tissue
Neural networks
Classifiers
image analysis
artificial intelligence
Pathology
Hematoxylin
Eosine Yellowish-(YS)
adenocarcinoma
processing technology
colorectal neoplasms
histopathology
Image analysis
Learning systems
Colorectal Neoplasms
Adenocarcinoma
Image processing

Fields of Expertise

  • Information, Communication & Computing

Cooperations

  • BioTechMed-Graz

Cite this

Segmentation and classification of colon glands with deep convolutional neural networks and total variation regularization. / Kainz, Philipp; Pfeiffer, Michael; Urschler, Martin.

In: PeerJ, Vol. 5, e3874, 03.10.2017.

Research output: Contribution to journalArticleResearchpeer-review

@article{29354a6cd5614503a826723028c4f068,
title = "Segmentation and classification of colon glands with deep convolutional neural networks and total variation regularization",
abstract = "Segmentation of histopathology sections is a necessary preprocessing step for digital pathology. Due to the large variability of biological tissue, machine learning techniques have shown superior performance over conventional image processing methods. Here we present our deep neural network-based approach for segmentation and classification of glands in tissue of benign and malignant colorectal cancer, which was developed to participate in the GlaS@MICCAI2015 colon gland segmentation challenge. We use two distinct deep convolutional neural networks (CNN) for pixel-wise classification of Hematoxylin-Eosin stained images. While the first classifier separates glands from background, the second classifier identifies gland-separating structures. In a subsequent step, a figure-ground segmentation based on weighted total variation produces the final segmentation result by regularizing the CNN predictions. We present both quantitative and qualitative segmentation results on the recently released and publicly available Warwick-QU colon adenocarcinoma dataset associated with the GlaS@MICCAI2015 challenge and compare our approach to the simultaneously developed other approaches that participated in the same challenge. On two test sets, we demonstrate our segmentation performance and show that we achieve a tissue classification accuracy of 98{\%} and 95{\%}, making use of the inherent capability of our system to distinguish between benign and malignant tissue. Our results show that deep learning approaches can yield highly accurate and reproducible results for biomedical image analysis, with the potential to significantly improve the quality and speed of medical diagnoses.",
author = "Philipp Kainz and Michael Pfeiffer and Martin Urschler",
year = "2017",
month = "10",
day = "3",
doi = "10.7717/peerj.3874",
language = "English",
volume = "5",
journal = "PeerJ",
issn = "2167-8359",
publisher = "PeerJ, Inc.",

}

TY - JOUR

T1 - Segmentation and classification of colon glands with deep convolutional neural networks and total variation regularization

AU - Kainz,Philipp

AU - Pfeiffer,Michael

AU - Urschler,Martin

PY - 2017/10/3

Y1 - 2017/10/3

N2 - Segmentation of histopathology sections is a necessary preprocessing step for digital pathology. Due to the large variability of biological tissue, machine learning techniques have shown superior performance over conventional image processing methods. Here we present our deep neural network-based approach for segmentation and classification of glands in tissue of benign and malignant colorectal cancer, which was developed to participate in the GlaS@MICCAI2015 colon gland segmentation challenge. We use two distinct deep convolutional neural networks (CNN) for pixel-wise classification of Hematoxylin-Eosin stained images. While the first classifier separates glands from background, the second classifier identifies gland-separating structures. In a subsequent step, a figure-ground segmentation based on weighted total variation produces the final segmentation result by regularizing the CNN predictions. We present both quantitative and qualitative segmentation results on the recently released and publicly available Warwick-QU colon adenocarcinoma dataset associated with the GlaS@MICCAI2015 challenge and compare our approach to the simultaneously developed other approaches that participated in the same challenge. On two test sets, we demonstrate our segmentation performance and show that we achieve a tissue classification accuracy of 98% and 95%, making use of the inherent capability of our system to distinguish between benign and malignant tissue. Our results show that deep learning approaches can yield highly accurate and reproducible results for biomedical image analysis, with the potential to significantly improve the quality and speed of medical diagnoses.

AB - Segmentation of histopathology sections is a necessary preprocessing step for digital pathology. Due to the large variability of biological tissue, machine learning techniques have shown superior performance over conventional image processing methods. Here we present our deep neural network-based approach for segmentation and classification of glands in tissue of benign and malignant colorectal cancer, which was developed to participate in the GlaS@MICCAI2015 colon gland segmentation challenge. We use two distinct deep convolutional neural networks (CNN) for pixel-wise classification of Hematoxylin-Eosin stained images. While the first classifier separates glands from background, the second classifier identifies gland-separating structures. In a subsequent step, a figure-ground segmentation based on weighted total variation produces the final segmentation result by regularizing the CNN predictions. We present both quantitative and qualitative segmentation results on the recently released and publicly available Warwick-QU colon adenocarcinoma dataset associated with the GlaS@MICCAI2015 challenge and compare our approach to the simultaneously developed other approaches that participated in the same challenge. On two test sets, we demonstrate our segmentation performance and show that we achieve a tissue classification accuracy of 98% and 95%, making use of the inherent capability of our system to distinguish between benign and malignant tissue. Our results show that deep learning approaches can yield highly accurate and reproducible results for biomedical image analysis, with the potential to significantly improve the quality and speed of medical diagnoses.

U2 - 10.7717/peerj.3874

DO - 10.7717/peerj.3874

M3 - Article

VL - 5

JO - PeerJ

T2 - PeerJ

JF - PeerJ

SN - 2167-8359

M1 - e3874

ER -