Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations

Hong Yu Zhou; Shuang Yu; Cheng Bian; Yifan Hu; Kai Ma; Yefeng Zheng

doi:10.1007/978-3-030-59710-8_39

Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations

Hong Yu Zhou
, Shuang Yu
, Cheng Bian^*
, Yifan Hu
, Kai Ma
, Yefeng Zheng

^*Corresponding author for this work

Tencent

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

94 Scopus citations

Abstract

In deep learning era, pretrained models play an important role in medical image analysis, in which ImageNet pretraining has been widely adopted as the best way. However, it is undeniable that there exists an obvious domain gap between natural images and medical images. To bridge this gap, we propose a new pretraining method which learns from 700k radiographs given no manual annotations. We call our method as Comparing to Learn (C2L) because it learns robust features by comparing different image representations. To verify the effectiveness of C2L, we conduct comprehensive ablation studies and evaluate it on different tasks and datasets. The experimental results on radiographs show that C2L can outperform ImageNet pretraining and previous state-of-the-art approaches significantly. Code and models are available at https://github.com/funnyzhou/C2L_MICCAI2020.

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings
Editors	Anne L. Martel, Purang Abolmaesumi, Danail Stoyanov, Diana Mateus, Maria A. Zuluaga, S. Kevin Zhou, Daniel Racoceanu, Leo Joskowicz
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	398-407
Number of pages	10
ISBN (Print)	9783030597092
DOIs	https://doi.org/10.1007/978-3-030-59710-8_39
State	Published - 2020
Externally published	Yes
Event	23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020 - Lima, Peru Duration: 4 Oct 2020 → 8 Oct 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12261 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020
Country/Territory	Peru
City	Lima
Period	4/10/20 → 8/10/20

Keywords

Pretrained models
Radiograph
Self-supervised learning

Access to Document

10.1007/978-3-030-59710-8_39

Cite this

Zhou, H. Y., Yu, S., Bian, C., Hu, Y., Ma, K., & Zheng, Y. (2020). Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations. In A. L. Martel, P. Abolmaesumi, D. Stoyanov, D. Mateus, M. A. Zuluaga, S. K. Zhou, D. Racoceanu, & L. Joskowicz (Eds.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings (pp. 398-407). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12261 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-59710-8_39

Zhou, Hong Yu ; Yu, Shuang ; Bian, Cheng et al. / Comparing to learn : Surpassing imagenet pretraining on radiographs by comparing image representations. Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings. editor / Anne L. Martel ; Purang Abolmaesumi ; Danail Stoyanov ; Diana Mateus ; Maria A. Zuluaga ; S. Kevin Zhou ; Daniel Racoceanu ; Leo Joskowicz. Springer Science and Business Media Deutschland GmbH, 2020. pp. 398-407 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{5d5a2e26f511464aa699bdb6441ef3a0,

title = "Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations",

abstract = "In deep learning era, pretrained models play an important role in medical image analysis, in which ImageNet pretraining has been widely adopted as the best way. However, it is undeniable that there exists an obvious domain gap between natural images and medical images. To bridge this gap, we propose a new pretraining method which learns from 700k radiographs given no manual annotations. We call our method as Comparing to Learn (C2L) because it learns robust features by comparing different image representations. To verify the effectiveness of C2L, we conduct comprehensive ablation studies and evaluate it on different tasks and datasets. The experimental results on radiographs show that C2L can outperform ImageNet pretraining and previous state-of-the-art approaches significantly. Code and models are available at https://github.com/funnyzhou/C2L\_MICCAI2020.",

keywords = "Pretrained models, Radiograph, Self-supervised learning",

author = "Zhou, \{Hong Yu\} and Shuang Yu and Cheng Bian and Yifan Hu and Kai Ma and Yefeng Zheng",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2020.; 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020 ; Conference date: 04-10-2020 Through 08-10-2020",

year = "2020",

doi = "10.1007/978-3-030-59710-8\_39",

language = "英语",

isbn = "9783030597092",

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 = "398--407",

editor = "Martel, \{Anne L.\} and Purang Abolmaesumi and Danail Stoyanov and Diana Mateus and Zuluaga, \{Maria A.\} and Zhou, \{S. Kevin\} and Daniel Racoceanu and Leo Joskowicz",

booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings",

address = "德国",

}

Zhou, HY, Yu, S, Bian, C, Hu, Y, Ma, K & Zheng, Y 2020, Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations. in AL Martel, P Abolmaesumi, D Stoyanov, D Mateus, MA Zuluaga, SK Zhou, D Racoceanu & L Joskowicz (eds), Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12261 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 398-407, 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020, Lima, Peru, 4/10/20. https://doi.org/10.1007/978-3-030-59710-8_39

Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations. / Zhou, Hong Yu; Yu, Shuang; Bian, Cheng et al.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings. ed. / Anne L. Martel; Purang Abolmaesumi; Danail Stoyanov; Diana Mateus; Maria A. Zuluaga; S. Kevin Zhou; Daniel Racoceanu; Leo Joskowicz. Springer Science and Business Media Deutschland GmbH, 2020. p. 398-407 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12261 LNCS).

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

TY - GEN

T1 - Comparing to learn

T2 - 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020

AU - Zhou, Hong Yu

AU - Yu, Shuang

AU - Bian, Cheng

AU - Hu, Yifan

AU - Ma, Kai

AU - Zheng, Yefeng

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2020.

PY - 2020

Y1 - 2020

N2 - In deep learning era, pretrained models play an important role in medical image analysis, in which ImageNet pretraining has been widely adopted as the best way. However, it is undeniable that there exists an obvious domain gap between natural images and medical images. To bridge this gap, we propose a new pretraining method which learns from 700k radiographs given no manual annotations. We call our method as Comparing to Learn (C2L) because it learns robust features by comparing different image representations. To verify the effectiveness of C2L, we conduct comprehensive ablation studies and evaluate it on different tasks and datasets. The experimental results on radiographs show that C2L can outperform ImageNet pretraining and previous state-of-the-art approaches significantly. Code and models are available at https://github.com/funnyzhou/C2L_MICCAI2020.

AB - In deep learning era, pretrained models play an important role in medical image analysis, in which ImageNet pretraining has been widely adopted as the best way. However, it is undeniable that there exists an obvious domain gap between natural images and medical images. To bridge this gap, we propose a new pretraining method which learns from 700k radiographs given no manual annotations. We call our method as Comparing to Learn (C2L) because it learns robust features by comparing different image representations. To verify the effectiveness of C2L, we conduct comprehensive ablation studies and evaluate it on different tasks and datasets. The experimental results on radiographs show that C2L can outperform ImageNet pretraining and previous state-of-the-art approaches significantly. Code and models are available at https://github.com/funnyzhou/C2L_MICCAI2020.

KW - Pretrained models

KW - Radiograph

KW - Self-supervised learning

UR - https://www.scopus.com/pages/publications/85093124168

U2 - 10.1007/978-3-030-59710-8_39

DO - 10.1007/978-3-030-59710-8_39

M3 - 会议稿件

AN - SCOPUS:85093124168

SN - 9783030597092

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

SP - 398

EP - 407

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings

A2 - Martel, Anne L.

A2 - Abolmaesumi, Purang

A2 - Stoyanov, Danail

A2 - Mateus, Diana

A2 - Zuluaga, Maria A.

A2 - Zhou, S. Kevin

A2 - Racoceanu, Daniel

A2 - Joskowicz, Leo

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 4 October 2020 through 8 October 2020

ER -

Zhou HY, Yu S, Bian C, Hu Y, Ma K, Zheng Y. Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations. In Martel AL, Abolmaesumi P, Stoyanov D, Mateus D, Zuluaga MA, Zhou SK, Racoceanu D, Joskowicz L, editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2020. p. 398-407. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-59710-8_39

Comparing to learn: Surpassing imagenet pretraining on radiographs by comparing image representations

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this