Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification

Pengfei Diao; Akshay Pai; Christian Igel; Christian Hedeager Krag

doi:10.1007/978-3-031-16449-1_72

Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification

Pengfei Diao, Akshay Pai^*, Christian Igel, Christian Hedeager Krag

^*Corresponding author for this work

4 Citations (Scopus)

Abstract

Domain shift is a common problem in machine learning and medical imaging. Currently one of the most popular domain adaptation approaches is the domain-invariant mapping method using generative adversarial networks (GANs). These methods deploy some variation of a GAN to learn target domain distributions which work on pixel level. However, they often produce too complicated or unnecessary transformations. This paper is based on the hypothesis that most domain shifts in medical images are variations of global intensity changes which can be captured by transforming histograms along with individual pixel intensities. We propose a histogram-based GAN methodology for domain adaptation that outperforms standard pixel-based GAN methods in classifying chest x-rays from various heterogeneous target domains.

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings
Editors	Linwei Wang, Qi Dou, P. Thomas Fletcher, Stefanie Speidel, Shuo Li
Number of pages	10
Publisher	Springer Science and Business Media Deutschland GmbH
Publication date	2022
Pages	755-764
ISBN (Print)	9783031164484
DOIs	https://doi.org/10.1007/978-3-031-16449-1_72
Publication status	Published - 2022
Event	25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022 - Singapore, Singapore Duration: 18 Sept 2022 → 22 Sept 2022

Conference

Conference	25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022
Country/Territory	Singapore
City	Singapore
Period	18/09/2022 → 22/09/2022

Series	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13437 LNCS
ISSN	0302-9743

Keywords

Histogram layer
Lung disease classification
Unsupervised domain adaptation

Access to Document

10.1007/978-3-031-16449-1_72

Cite this

Diao, P., Pai, A., Igel, C., & Krag, C. H. (2022). Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification. In L. Wang, Q. Dou, P. T. Fletcher, S. Speidel, & S. Li (Eds.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings (pp. 755-764). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-16449-1_72

Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification. / Diao, Pengfei; Pai, Akshay; Igel, Christian et al.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. ed. / Linwei Wang; Qi Dou; P. Thomas Fletcher; Stefanie Speidel; Shuo Li. Springer Science and Business Media Deutschland GmbH, 2022. p. 755-764 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13437 LNCS).

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

Diao, P, Pai, A, Igel, C & Krag, CH 2022, Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification. in L Wang, Q Dou, PT Fletcher, S Speidel & S Li (eds), Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. Springer Science and Business Media Deutschland GmbH, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13437 LNCS, pp. 755-764, 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022, Singapore, Singapore, 18/09/2022. https://doi.org/10.1007/978-3-031-16449-1_72

Diao P, Pai A, Igel C, Krag CH. Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification. In Wang L, Dou Q, Fletcher PT, Speidel S, Li S, editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2022. p. 755-764. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13437 LNCS). doi: 10.1007/978-3-031-16449-1_72

Diao, Pengfei ; Pai, Akshay ; Igel, Christian et al. / Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification. Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. editor / Linwei Wang ; Qi Dou ; P. Thomas Fletcher ; Stefanie Speidel ; Shuo Li. Springer Science and Business Media Deutschland GmbH, 2022. pp. 755-764 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13437 LNCS).

@inproceedings{cb16842f383a425c91fa75557c2057c1,

title = "Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification",

abstract = "Domain shift is a common problem in machine learning and medical imaging. Currently one of the most popular domain adaptation approaches is the domain-invariant mapping method using generative adversarial networks (GANs). These methods deploy some variation of a GAN to learn target domain distributions which work on pixel level. However, they often produce too complicated or unnecessary transformations. This paper is based on the hypothesis that most domain shifts in medical images are variations of global intensity changes which can be captured by transforming histograms along with individual pixel intensities. We propose a histogram-based GAN methodology for domain adaptation that outperforms standard pixel-based GAN methods in classifying chest x-rays from various heterogeneous target domains.",

keywords = "Histogram layer, Lung disease classification, Unsupervised domain adaptation",

author = "Pengfei Diao and Akshay Pai and Christian Igel and Krag, {Christian Hedeager}",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022 ; Conference date: 18-09-2022 Through 22-09-2022",

year = "2022",

doi = "10.1007/978-3-031-16449-1_72",

language = "English",

isbn = "9783031164484",

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 = "755--764",

editor = "Linwei Wang and Qi Dou and Fletcher, {P. Thomas} and Stefanie Speidel and Shuo Li",

booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings",

address = "Germany",

}

TY - GEN

T1 - Histogram-Based Unsupervised Domain Adaptation for Medical Image Classification

AU - Diao, Pengfei

AU - Pai, Akshay

AU - Igel, Christian

AU - Krag, Christian Hedeager

PY - 2022

Y1 - 2022

N2 - Domain shift is a common problem in machine learning and medical imaging. Currently one of the most popular domain adaptation approaches is the domain-invariant mapping method using generative adversarial networks (GANs). These methods deploy some variation of a GAN to learn target domain distributions which work on pixel level. However, they often produce too complicated or unnecessary transformations. This paper is based on the hypothesis that most domain shifts in medical images are variations of global intensity changes which can be captured by transforming histograms along with individual pixel intensities. We propose a histogram-based GAN methodology for domain adaptation that outperforms standard pixel-based GAN methods in classifying chest x-rays from various heterogeneous target domains.

AB - Domain shift is a common problem in machine learning and medical imaging. Currently one of the most popular domain adaptation approaches is the domain-invariant mapping method using generative adversarial networks (GANs). These methods deploy some variation of a GAN to learn target domain distributions which work on pixel level. However, they often produce too complicated or unnecessary transformations. This paper is based on the hypothesis that most domain shifts in medical images are variations of global intensity changes which can be captured by transforming histograms along with individual pixel intensities. We propose a histogram-based GAN methodology for domain adaptation that outperforms standard pixel-based GAN methods in classifying chest x-rays from various heterogeneous target domains.

KW - Histogram layer

KW - Lung disease classification

KW - Unsupervised domain adaptation

UR - http://www.scopus.com/inward/record.url?scp=85139020088&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-16449-1_72

DO - 10.1007/978-3-031-16449-1_72

M3 - Article in proceedings

AN - SCOPUS:85139020088

SN - 9783031164484

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

SP - 755

EP - 764

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings

A2 - Wang, Linwei

A2 - Dou, Qi

A2 - Fletcher, P. Thomas

A2 - Speidel, Stefanie

A2 - Li, Shuo

PB - Springer Science and Business Media Deutschland GmbH

T2 - 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022

Y2 - 18 September 2022 through 22 September 2022

ER -