Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection

Saad Ullah Akram; Talha Qaiser; Simon Graham; Juho Kannala; Janne Heikkilä; Nasir Rajpoot

doi:10.1007/978-3-030-00949-6_9

Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection

Saad Ullah Akram^*, Talha Qaiser, Simon Graham, Juho Kannala, Janne Heikkilä, Nasir Rajpoot

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

26 Citations (Scopus)

Abstract

Mitosis count is an important biomarker for prognosis of various cancers. At present, pathologists typically perform manual counting on a few selected regions of interest in breast whole-slide-images (WSIs) of patient biopsies. This task is very time-consuming, tedious and subjective. Automated mitosis detection methods have made great advances in recent years. However, these methods require exhaustive labeling of a large number of selected regions of interest. This task is very expensive because expert pathologists are needed for reliable and accurate annotations. In this paper, we present a semi-supervised mitosis detection method which is designed to leverage a large number of unlabeled breast cancer WSIs. As a result, our method capitalizes on the growing number of digitized histology images, without relying on exhaustive annotations, subsequently improving mitosis detection. Our method first learns a mitosis detector from labeled data, uses this detector to mine additional mitosis samples from unlabeled WSIs, and then trains the final model using this larger and diverse set of mitosis samples. The use of unlabeled data improves F1-score by ∼ 5% compared to our best performing fully-supervised model on the TUPAC validation set. Our submission (single model) to TUPAC challenge ranks highly on the leaderboard with an F1-score of 0.64.

Original language	English
Title of host publication	Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings
Publisher	Springer
Pages	69-77
Number of pages	9
ISBN (Electronic)	9783030009496
ISBN (Print)	9783030009489
DOIs	https://doi.org/10.1007/978-3-030-00949-6_9
Publication status	Published - 1 Jan 2018
MoE publication type	A4 Conference publication
Event	International Workshop on Computational Pathology - Granada, Spain Duration: 16 Sept 2018 → 20 Sept 2018 Conference number: 1

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer
Volume	11039 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Workshop

Workshop	International Workshop on Computational Pathology
Abbreviated title	COMPAY
Country/Territory	Spain
City	Granada
Period	16/09/2018 → 20/09/2018

Keywords

Breast cancer
Computational pathology
Mitosis detection
Self-supervised learning
Semi-supervised learning

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-030-00949-6_9

https://arxiv.org/abs/1807.11677

Cite this

Akram, S. U., Qaiser, T., Graham, S., Kannala, J., Heikkilä, J., & Rajpoot, N. (2018). Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection. In Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings (pp. 69-77). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11039 LNCS). Springer. https://doi.org/10.1007/978-3-030-00949-6_9

Akram, Saad Ullah ; Qaiser, Talha ; Graham, Simon et al. / Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection. Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. Springer, 2018. pp. 69-77 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{efc75720bf6b45f69ae0e0168d57a1cb,

title = "Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection",

abstract = "Mitosis count is an important biomarker for prognosis of various cancers. At present, pathologists typically perform manual counting on a few selected regions of interest in breast whole-slide-images (WSIs) of patient biopsies. This task is very time-consuming, tedious and subjective. Automated mitosis detection methods have made great advances in recent years. However, these methods require exhaustive labeling of a large number of selected regions of interest. This task is very expensive because expert pathologists are needed for reliable and accurate annotations. In this paper, we present a semi-supervised mitosis detection method which is designed to leverage a large number of unlabeled breast cancer WSIs. As a result, our method capitalizes on the growing number of digitized histology images, without relying on exhaustive annotations, subsequently improving mitosis detection. Our method first learns a mitosis detector from labeled data, uses this detector to mine additional mitosis samples from unlabeled WSIs, and then trains the final model using this larger and diverse set of mitosis samples. The use of unlabeled data improves F1-score by ∼ 5% compared to our best performing fully-supervised model on the TUPAC validation set. Our submission (single model) to TUPAC challenge ranks highly on the leaderboard with an F1-score of 0.64.",

keywords = "Breast cancer, Computational pathology, Mitosis detection, Self-supervised learning, Semi-supervised learning",

author = "Akram, {Saad Ullah} and Talha Qaiser and Simon Graham and Juho Kannala and Janne Heikkil{\"a} and Nasir Rajpoot",

year = "2018",

month = jan,

day = "1",

doi = "10.1007/978-3-030-00949-6_9",

language = "English",

isbn = "9783030009489",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "69--77",

booktitle = "Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings",

address = "Germany",

note = "International Workshop on Computational Pathology, COMPAY ; Conference date: 16-09-2018 Through 20-09-2018",

}

Akram, SU, Qaiser, T, Graham, S, Kannala, J, Heikkilä, J & Rajpoot, N 2018, Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection. in Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11039 LNCS, Springer, pp. 69-77, International Workshop on Computational Pathology, Granada, Spain, 16/09/2018. https://doi.org/10.1007/978-3-030-00949-6_9

Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection. / Akram, Saad Ullah; Qaiser, Talha; Graham, Simon et al.
Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. Springer, 2018. p. 69-77 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11039 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection

AU - Akram, Saad Ullah

AU - Qaiser, Talha

AU - Graham, Simon

AU - Kannala, Juho

AU - Heikkilä, Janne

AU - Rajpoot, Nasir

N1 - Conference code: 1

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Mitosis count is an important biomarker for prognosis of various cancers. At present, pathologists typically perform manual counting on a few selected regions of interest in breast whole-slide-images (WSIs) of patient biopsies. This task is very time-consuming, tedious and subjective. Automated mitosis detection methods have made great advances in recent years. However, these methods require exhaustive labeling of a large number of selected regions of interest. This task is very expensive because expert pathologists are needed for reliable and accurate annotations. In this paper, we present a semi-supervised mitosis detection method which is designed to leverage a large number of unlabeled breast cancer WSIs. As a result, our method capitalizes on the growing number of digitized histology images, without relying on exhaustive annotations, subsequently improving mitosis detection. Our method first learns a mitosis detector from labeled data, uses this detector to mine additional mitosis samples from unlabeled WSIs, and then trains the final model using this larger and diverse set of mitosis samples. The use of unlabeled data improves F1-score by ∼ 5% compared to our best performing fully-supervised model on the TUPAC validation set. Our submission (single model) to TUPAC challenge ranks highly on the leaderboard with an F1-score of 0.64.

AB - Mitosis count is an important biomarker for prognosis of various cancers. At present, pathologists typically perform manual counting on a few selected regions of interest in breast whole-slide-images (WSIs) of patient biopsies. This task is very time-consuming, tedious and subjective. Automated mitosis detection methods have made great advances in recent years. However, these methods require exhaustive labeling of a large number of selected regions of interest. This task is very expensive because expert pathologists are needed for reliable and accurate annotations. In this paper, we present a semi-supervised mitosis detection method which is designed to leverage a large number of unlabeled breast cancer WSIs. As a result, our method capitalizes on the growing number of digitized histology images, without relying on exhaustive annotations, subsequently improving mitosis detection. Our method first learns a mitosis detector from labeled data, uses this detector to mine additional mitosis samples from unlabeled WSIs, and then trains the final model using this larger and diverse set of mitosis samples. The use of unlabeled data improves F1-score by ∼ 5% compared to our best performing fully-supervised model on the TUPAC validation set. Our submission (single model) to TUPAC challenge ranks highly on the leaderboard with an F1-score of 0.64.

KW - Breast cancer

KW - Computational pathology

KW - Mitosis detection

KW - Self-supervised learning

KW - Semi-supervised learning

UR - http://www.scopus.com/inward/record.url?scp=85053863824&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-00949-6_9

DO - 10.1007/978-3-030-00949-6_9

M3 - Conference article in proceedings

AN - SCOPUS:85053863824

SN - 9783030009489

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

SP - 69

EP - 77

BT - Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings

PB - Springer

T2 - International Workshop on Computational Pathology

Y2 - 16 September 2018 through 20 September 2018

ER -

Akram SU, Qaiser T, Graham S, Kannala J, Heikkilä J, Rajpoot N. Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection. In Computational Pathology and Ophthalmic Medical Image Analysis - First International Workshop, COMPAY 2018, and 5th International Workshop, OMIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. Springer. 2018. p. 69-77. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-00949-6_9

Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection

Abstract

Publication series

Workshop

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this