Iterative learning to make the most of unlabeled and quickly obtained labeled data in histology

L. Gupta; B.M. Klinkhammer; P. Boor; D. Merhof; M. Gadermayr

Iterative learning to make the most of unlabeled and quickly obtained labeled data in histology

L. Gupta
, B.M. Klinkhammer
, P. Boor
, D. Merhof
, M. Gadermayr

Institute of Imaging & Computer Vision, RWTH Aachen University
Institute of Pathology, RWTH Aachen University Hospital

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Due to the increasing availability of digital whole slide scanners, the importance of image analysis in the field of digital pathology increased significantly. A major challenge and an equally big opportunity for analyses in this field is given by the wide range of tasks and different histological stains. Although sufficient image data is often available for training, the requirement for corresponding expert annotations inhibits clinical deployment. Thus, there is an urgent need for methods which can be effectively trained with or adapted to a small amount of labeled training data. Here, we propose a method to find an optimum trade-off between (low) annotation effort and (high) segmentation accuracy. For this purpose, we propose an approach based on a weakly supervised and an unsupervised learning stage relying on few roughly labeled samples and many unlabeled samples. Although the idea of weakly annotated data is not new, we firstly investigate the applicability to digital pathology in a state-of-the-art machine learning setting. © 2019 L. Gupta, B. Mara Klinkhammer, P. Boor, D. Merhof & M. Gadermayr.

Original language	English
Title of host publication	Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning
Pages	215-224
Number of pages	10
Volume	102
Publication status	Published - 2019
Event	2nd International Conference on Medical Imaging with Deep Learning, MIDL 2019 - London, United Kingdom Duration: 8 Jul 2019 → 10 Jul 2019 https://2019.midl.io/

Conference

Conference	2nd International Conference on Medical Imaging with Deep Learning, MIDL 2019
Abbreviated title	MIDL 2019
Country/Territory	United Kingdom
City	London
Period	8/07/19 → 10/07/19
Internet address	https://2019.midl.io/

Keywords

convolutional neural networks
Digital pathology
kidney
segmentation
weakly supervised
E-learning
Economic and social effects
Learning algorithms
Machine learning
Neural networks
Pathology
Convolutional neural network
Digital pathologies
Expert annotations
Image data
Image-analysis
Iterative learning
Kidney
Labeled data
Segmentation
Weakly supervised
Iterative methods

Access to Document

Cite this

@inproceedings{6b0eb4136db04f4ab7c0e39f895eed42,

title = "Iterative learning to make the most of unlabeled and quickly obtained labeled data in histology",

abstract = "Due to the increasing availability of digital whole slide scanners, the importance of image analysis in the field of digital pathology increased significantly. A major challenge and an equally big opportunity for analyses in this field is given by the wide range of tasks and different histological stains. Although sufficient image data is often available for training, the requirement for corresponding expert annotations inhibits clinical deployment. Thus, there is an urgent need for methods which can be effectively trained with or adapted to a small amount of labeled training data. Here, we propose a method to find an optimum trade-off between (low) annotation effort and (high) segmentation accuracy. For this purpose, we propose an approach based on a weakly supervised and an unsupervised learning stage relying on few roughly labeled samples and many unlabeled samples. Although the idea of weakly annotated data is not new, we firstly investigate the applicability to digital pathology in a state-of-the-art machine learning setting. {\textcopyright} 2019 L. Gupta, B. Mara Klinkhammer, P. Boor, D. Merhof \& M. Gadermayr.",

keywords = "convolutional neural networks, Digital pathology, kidney, segmentation, weakly supervised, E-learning, Economic and social effects, Learning algorithms, Machine learning, Neural networks, Pathology, Convolutional neural network, Digital pathologies, Expert annotations, Image data, Image-analysis, Iterative learning, Kidney, Labeled data, Segmentation, Weakly supervised, Iterative methods",

author = "L. Gupta and B.M. Klinkhammer and P. Boor and D. Merhof and M. Gadermayr",

note = "Conference code: 188845 Cited By :6 Export Date: 14 December 2023 Funding details: Deutsche Forschungsgemeinschaft, DFG, ME3737/3-1 Funding text 1: This work was supported by the German Research Foundation (DFG) under grant no. ME3737/3-1. References: Gadermayr, Michael, Appel, Vitus, Klinkhammer, Barbara M., Boor, Peter, Merhof, Dorit, Which way round? a study on the performance of stain-translation for segmenting arbitrarily dyed histological images (2018) Proceedings of the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI'18); Gadermayr, Michael, Dombrowski, Ann-Kathrin, Klinkhammer, Barbara Mara, Boor, Peter, Merhof, Dorit, Cnn cascades for segmenting sparse objects in gigapixel whole slide images (2019) Computerized Medical Imaging and Graphics; Herve, N., Servais, A., Thervet, E., Olivo-Marin, J.-C., Meas-Yedid, V., Statistical color texture descriptors for histological images analysis (2011) Proceedings of ISBI'11, pp. 724-727; Kato, Tsuyoshi, Relator, Raissa, Ngouv, Hayliang, Hirohashi, Yoshihiro, Takaki, Osamu, Kakimoto, Tetsuhiro, Okada, Kinya, Segmental HOG: new descriptor for glomerulus detection in kidney microscopy image (2015) BMC Bioinformatics, 16 (1); Khoreva, Anna, Benenson, Rodrigo, Hosang, Jan, Hein, Matthias, Schiele, Bernt, Simple does it: Weakly supervised instance and semantic segmentation (2017) Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR'17); Komura, Daisuke, Ishikawa, Shumpei, Machine learning methods for histopathological image analysis (2018) Computational and Structural Biotechnology Journal, 16, pp. 34-42; Ronneberger, Olaf, Fischer, Philipp, Brox, Thomas, U-net: Convolutional networks for biomedical image segmentation (2015) Proceedings of the International Conference on Medical Image Computing and Computer Aided Interventions (MICCAI'15), pp. 234-241. , Springer International Publishing; 2nd International Conference on Medical Imaging with Deep Learning, MIDL 2019, MIDL 2019 ; Conference date: 08-07-2019 Through 10-07-2019",

year = "2019",

language = "English",

volume = "102",

pages = "215--224",

booktitle = "Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning",

url = "https://2019.midl.io/",

}

Gupta, L, Klinkhammer, BM, Boor, P, Merhof, D & Gadermayr, M 2019, Iterative learning to make the most of unlabeled and quickly obtained labeled data in histology. in Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning. vol. 102, pp. 215-224, 2nd International Conference on Medical Imaging with Deep Learning, MIDL 2019, London, United Kingdom, 8/07/19. <https://proceedings.mlr.press/v102/gupta19a.html>

TY - GEN

T1 - Iterative learning to make the most of unlabeled and quickly obtained labeled data in histology

AU - Gupta, L.

AU - Klinkhammer, B.M.

AU - Boor, P.

AU - Merhof, D.

AU - Gadermayr, M.

N1 - Conference code: 188845 Cited By :6 Export Date: 14 December 2023 Funding details: Deutsche Forschungsgemeinschaft, DFG, ME3737/3-1 Funding text 1: This work was supported by the German Research Foundation (DFG) under grant no. ME3737/3-1. References: Gadermayr, Michael, Appel, Vitus, Klinkhammer, Barbara M., Boor, Peter, Merhof, Dorit, Which way round? a study on the performance of stain-translation for segmenting arbitrarily dyed histological images (2018) Proceedings of the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI'18); Gadermayr, Michael, Dombrowski, Ann-Kathrin, Klinkhammer, Barbara Mara, Boor, Peter, Merhof, Dorit, Cnn cascades for segmenting sparse objects in gigapixel whole slide images (2019) Computerized Medical Imaging and Graphics; Herve, N., Servais, A., Thervet, E., Olivo-Marin, J.-C., Meas-Yedid, V., Statistical color texture descriptors for histological images analysis (2011) Proceedings of ISBI'11, pp. 724-727; Kato, Tsuyoshi, Relator, Raissa, Ngouv, Hayliang, Hirohashi, Yoshihiro, Takaki, Osamu, Kakimoto, Tetsuhiro, Okada, Kinya, Segmental HOG: new descriptor for glomerulus detection in kidney microscopy image (2015) BMC Bioinformatics, 16 (1); Khoreva, Anna, Benenson, Rodrigo, Hosang, Jan, Hein, Matthias, Schiele, Bernt, Simple does it: Weakly supervised instance and semantic segmentation (2017) Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR'17); Komura, Daisuke, Ishikawa, Shumpei, Machine learning methods for histopathological image analysis (2018) Computational and Structural Biotechnology Journal, 16, pp. 34-42; Ronneberger, Olaf, Fischer, Philipp, Brox, Thomas, U-net: Convolutional networks for biomedical image segmentation (2015) Proceedings of the International Conference on Medical Image Computing and Computer Aided Interventions (MICCAI'15), pp. 234-241. , Springer International Publishing

PY - 2019

Y1 - 2019

N2 - Due to the increasing availability of digital whole slide scanners, the importance of image analysis in the field of digital pathology increased significantly. A major challenge and an equally big opportunity for analyses in this field is given by the wide range of tasks and different histological stains. Although sufficient image data is often available for training, the requirement for corresponding expert annotations inhibits clinical deployment. Thus, there is an urgent need for methods which can be effectively trained with or adapted to a small amount of labeled training data. Here, we propose a method to find an optimum trade-off between (low) annotation effort and (high) segmentation accuracy. For this purpose, we propose an approach based on a weakly supervised and an unsupervised learning stage relying on few roughly labeled samples and many unlabeled samples. Although the idea of weakly annotated data is not new, we firstly investigate the applicability to digital pathology in a state-of-the-art machine learning setting. © 2019 L. Gupta, B. Mara Klinkhammer, P. Boor, D. Merhof & M. Gadermayr.

AB - Due to the increasing availability of digital whole slide scanners, the importance of image analysis in the field of digital pathology increased significantly. A major challenge and an equally big opportunity for analyses in this field is given by the wide range of tasks and different histological stains. Although sufficient image data is often available for training, the requirement for corresponding expert annotations inhibits clinical deployment. Thus, there is an urgent need for methods which can be effectively trained with or adapted to a small amount of labeled training data. Here, we propose a method to find an optimum trade-off between (low) annotation effort and (high) segmentation accuracy. For this purpose, we propose an approach based on a weakly supervised and an unsupervised learning stage relying on few roughly labeled samples and many unlabeled samples. Although the idea of weakly annotated data is not new, we firstly investigate the applicability to digital pathology in a state-of-the-art machine learning setting. © 2019 L. Gupta, B. Mara Klinkhammer, P. Boor, D. Merhof & M. Gadermayr.

KW - convolutional neural networks

KW - Digital pathology

KW - kidney

KW - segmentation

KW - weakly supervised

KW - E-learning

KW - Economic and social effects

KW - Learning algorithms

KW - Machine learning

KW - Neural networks

KW - Pathology

KW - Convolutional neural network

KW - Digital pathologies

KW - Expert annotations

KW - Image data

KW - Image-analysis

KW - Iterative learning

KW - Kidney

KW - Labeled data

KW - Segmentation

KW - Weakly supervised

KW - Iterative methods

M3 - Conference contribution

VL - 102

SP - 215

EP - 224

BT - Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning

T2 - 2nd International Conference on Medical Imaging with Deep Learning, MIDL 2019

Y2 - 8 July 2019 through 10 July 2019

ER -

Iterative learning to make the most of unlabeled and quickly obtained labeled data in histology

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this