Hard Negative Sample Mining for Whole Slide Image Classification

Wentao Huang; Xiaoling Hu; Shahira Abousamra; Prateek Prasanna; Chao Chen

doi:10.1007/978-3-031-72083-3_14

Hard Negative Sample Mining for Whole Slide Image Classification

Wentao Huang
, Xiaoling Hu
, Shahira Abousamra
, Prateek Prasanna
, Chao Chen

Biomedical Informatics

Stony Brook University
Harvard University

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

5 Scopus citations

Abstract

Weakly supervised whole slide image (WSI) classification is challenging due to the lack of patch-level labels and high computational costs. State-of-the-art methods use self-supervised patch-wise feature representations for multiple instance learning (MIL). Recently, methods have been proposed to fine-tune the feature representation on the downstream task using pseudo labeling, but mostly focusing on selecting high-quality positive patches. In this paper, we propose to mine hard negative samples during fine-tuning. This allows us to obtain better feature representations and reduce the training cost. Furthermore, we propose a novel patch-wise ranking loss in MIL to better exploit these hard negative samples. Experiments on two public datasets demonstrate the efficacy of these proposed ideas. Our codes are available at https://github.com/winston52/HNM-WSI.

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings
Editors	Marius George Linguraru, Qi Dou, Aasa Feragen, Stamatia Giannarou, Ben Glocker, Karim Lekadir, Julia A. Schnabel
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	144-154
Number of pages	11
ISBN (Print)	9783031720826
DOIs	https://doi.org/10.1007/978-3-031-72083-3_14
State	Published - 2024
Event	27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024 - Marrakesh, Morocco Duration: Oct 6 2024 → Oct 10 2024

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	15004 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024
Country/Territory	Morocco
City	Marrakesh
Period	10/6/24 → 10/10/24

Keywords

Hard Sample Mining
Self-Training
Whole Slide Image

Access to Document

10.1007/978-3-031-72083-3_14

Cite this

Huang, W., Hu, X., Abousamra, S., Prasanna, P., & Chen, C. (2024). Hard Negative Sample Mining for Whole Slide Image Classification. In M. G. Linguraru, Q. Dou, A. Feragen, S. Giannarou, B. Glocker, K. Lekadir, & J. A. Schnabel (Eds.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings (pp. 144-154). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 15004 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-72083-3_14

Huang, Wentao ; Hu, Xiaoling ; Abousamra, Shahira et al. / Hard Negative Sample Mining for Whole Slide Image Classification. Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings. editor / Marius George Linguraru ; Qi Dou ; Aasa Feragen ; Stamatia Giannarou ; Ben Glocker ; Karim Lekadir ; Julia A. Schnabel. Springer Science and Business Media Deutschland GmbH, 2024. pp. 144-154 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{5368e2f912064394915340b697adbd8d,

title = "Hard Negative Sample Mining for Whole Slide Image Classification",

abstract = "Weakly supervised whole slide image (WSI) classification is challenging due to the lack of patch-level labels and high computational costs. State-of-the-art methods use self-supervised patch-wise feature representations for multiple instance learning (MIL). Recently, methods have been proposed to fine-tune the feature representation on the downstream task using pseudo labeling, but mostly focusing on selecting high-quality positive patches. In this paper, we propose to mine hard negative samples during fine-tuning. This allows us to obtain better feature representations and reduce the training cost. Furthermore, we propose a novel patch-wise ranking loss in MIL to better exploit these hard negative samples. Experiments on two public datasets demonstrate the efficacy of these proposed ideas. Our codes are available at https://github.com/winston52/HNM-WSI.",

keywords = "Hard Sample Mining, Self-Training, Whole Slide Image",

author = "Wentao Huang and Xiaoling Hu and Shahira Abousamra and Prateek Prasanna and Chao Chen",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.; 27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024 ; Conference date: 06-10-2024 Through 10-10-2024",

year = "2024",

doi = "10.1007/978-3-031-72083-3\_14",

language = "English",

isbn = "9783031720826",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "144--154",

editor = "Linguraru, \{Marius George\} and Qi Dou and Aasa Feragen and Stamatia Giannarou and Ben Glocker and Karim Lekadir and Schnabel, \{Julia A.\}",

booktitle = "Medical Image Computing and Computer Assisted Intervention {\textendash} MICCAI 2024 - 27th International Conference, Proceedings",

}

Huang, W, Hu, X, Abousamra, S, Prasanna, P & Chen, C 2024, Hard Negative Sample Mining for Whole Slide Image Classification. in MG Linguraru, Q Dou, A Feragen, S Giannarou, B Glocker, K Lekadir & JA Schnabel (eds), Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 15004 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 144-154, 27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024, Marrakesh, Morocco, 10/6/24. https://doi.org/10.1007/978-3-031-72083-3_14

Hard Negative Sample Mining for Whole Slide Image Classification. / Huang, Wentao; Hu, Xiaoling; Abousamra, Shahira et al.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings. ed. / Marius George Linguraru; Qi Dou; Aasa Feragen; Stamatia Giannarou; Ben Glocker; Karim Lekadir; Julia A. Schnabel. Springer Science and Business Media Deutschland GmbH, 2024. p. 144-154 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 15004 LNCS).

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

TY - GEN

T1 - Hard Negative Sample Mining for Whole Slide Image Classification

AU - Huang, Wentao

AU - Hu, Xiaoling

AU - Abousamra, Shahira

AU - Prasanna, Prateek

AU - Chen, Chao

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

PY - 2024

Y1 - 2024

N2 - Weakly supervised whole slide image (WSI) classification is challenging due to the lack of patch-level labels and high computational costs. State-of-the-art methods use self-supervised patch-wise feature representations for multiple instance learning (MIL). Recently, methods have been proposed to fine-tune the feature representation on the downstream task using pseudo labeling, but mostly focusing on selecting high-quality positive patches. In this paper, we propose to mine hard negative samples during fine-tuning. This allows us to obtain better feature representations and reduce the training cost. Furthermore, we propose a novel patch-wise ranking loss in MIL to better exploit these hard negative samples. Experiments on two public datasets demonstrate the efficacy of these proposed ideas. Our codes are available at https://github.com/winston52/HNM-WSI.

AB - Weakly supervised whole slide image (WSI) classification is challenging due to the lack of patch-level labels and high computational costs. State-of-the-art methods use self-supervised patch-wise feature representations for multiple instance learning (MIL). Recently, methods have been proposed to fine-tune the feature representation on the downstream task using pseudo labeling, but mostly focusing on selecting high-quality positive patches. In this paper, we propose to mine hard negative samples during fine-tuning. This allows us to obtain better feature representations and reduce the training cost. Furthermore, we propose a novel patch-wise ranking loss in MIL to better exploit these hard negative samples. Experiments on two public datasets demonstrate the efficacy of these proposed ideas. Our codes are available at https://github.com/winston52/HNM-WSI.

KW - Hard Sample Mining

KW - Self-Training

KW - Whole Slide Image

UR - https://www.scopus.com/pages/publications/85207642623

U2 - 10.1007/978-3-031-72083-3_14

DO - 10.1007/978-3-031-72083-3_14

M3 - Conference contribution

AN - SCOPUS:85207642623

SN - 9783031720826

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

SP - 144

EP - 154

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings

A2 - Linguraru, Marius George

A2 - Dou, Qi

A2 - Feragen, Aasa

A2 - Giannarou, Stamatia

A2 - Glocker, Ben

A2 - Lekadir, Karim

A2 - Schnabel, Julia A.

PB - Springer Science and Business Media Deutschland GmbH

T2 - 27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024

Y2 - 6 October 2024 through 10 October 2024

ER -

Huang W, Hu X, Abousamra S, Prasanna P , Chen C. Hard Negative Sample Mining for Whole Slide Image Classification. In Linguraru MG, Dou Q, Feragen A, Giannarou S, Glocker B, Lekadir K, Schnabel JA, editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 - 27th International Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2024. p. 144-154. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-72083-3_14