A Study on Domain Generalization with Data and Feature Augmentation for Robust Medical Image Segmentation = 견고한 의료 영상 구역화를 위한 데이터와 피쳐 증강 기반의 도메인 일반화 기법 연구

다국어 초록 (Multilingual Abstract)

Domain shift is a major challenge in deploying deep learning models for medical image segmentation across clinical settings. Deep learning models, trained exclusively on a single source domain, often struggle to generalize effectively to unseen target domains due to differences in imaging protocols, scanner vendors, and patient populations. For example, differences between camera specifications and lighting conditions can lead to significant discrepancies between the source and target domains. To address this significant issue, we propose two novel strategies for robust domain generalization: two illumination-based data augmentation methods inspired by the Retinex theory and a feature augmentation method that applies adaptive spectral random convolution to the feature space.
The proposed illumination-based data augmentation methods separate color medical images into two distinct components: illumination and reflectance. The first method is designed for medical images with relatively uniform and stable illumination conditions, such as fundus images. This method augments the overall illumination component by randomizing it to mimic variations in lighting conditions, thereby helping the segmentation model to generalize across unseen domains. The second method addresses medical images with more variable and complex illumination patterns, such as colonoscopy images. This method further decomposes the illumination component into global and local illumination components. A global illumination augmentation is performed to synthesize diverse training samples that more effectively capture the characteristics of unseen target domains.
Unlike the proposed illumination-based methods that focus on data augmentation in the image space, our proposed random convolution-based feature augmentation method, adaptive spectral random convolution (ASRConv), operates directly in the feature space, applying appropriate perturbations to the source domain features in the frequency domain without introducing high-frequency artifacts, encouraging the model to learn illumination-invariant feature representations. Extensive experiments conducted on multiple medical image segmentation datasets, including fundus and colonoscopy image datasets, demonstrate significant improvements in generalization performance on unseen target domains. The data augmentation-based methods outperform state-of-the-art domain generalization techniques by up to 9.60% in Dice scores, while the feature augmentation method improves existing feature augmentation-based domain generalization approaches by 3.07% in Dice scores.
The main contributions of this dissertation are threefold: (1) Develop two illumination-based data augmentation approaches that leverage the Retinex theory. (2) Propose a feature augmentation method, ASRConv, for domain generalization. (3) Conduct extensive experimental validation across diverse medical image datasets to confirm the efficacy of the proposed methods. By enhancing the robustness of deep learning models to domain shifts in medical images, these advancements can potentially improve the practicality of automated medical image analysis systems in real-world clinical settings, ultimately assisting clinicians in making more informed decisions by providing more accurate and consistent image segmentation results across different imaging conditions.

국문 초록 (Abstract)

도메인 전이(domain shift)는 다양한 임상 환경에서 의료 영상 영역화(segmentation)를 위한 딥러닝 모델을 적용할 때 정확도를 보장하기 위한 주된 도전 과제이다. 단일 원본 도메인에서 배타적으로 훈련된 딥러닝 모델은 종종 이기종의 영상 획득 프로토콜, 스캐너 제조사, 환자 특성의 차이로 인해 데이터 도메인의 변화가 발생하고 보이지 않는 목표 도메인에 효과적으로 일반화하는 데 어려움을 겪는다. 예를 들어, 카메라 사양과 조명 조건의 차이는 원본 도메인과 목표 도메인 간에 차이를 유발할 수 있으며 이로 인해 딥러닝 결과의 정확도가 낮아진다. 이러한 문제를 해결하기 위해 본 논문에서는 견고한 도메인 일반화를 위한 두 가지 기법을 제안한다. 첫째는 레티넥스(Retinex) 이론에 기반한 조명 관련 데이터 증강 방법이고 둘째는 적응형 스펙트럼 무작위 컨볼루션을 적용한 특징 공간 증강 방법을 제안한다.
제안된 조명 기반 데이터 증강 방법은 컬러 의료 영상에 대해 조명과 반사율 두 가지 구성 요소로 분리하여 처리 수준을 2단계로 구분한다. 우선 상대적으로 균일하고 안정된 조명 조건을 가진 의료 영상에 대한 처리로 망막 이미지 등과 같이 안정적이고 균일한 조명 조건을 가진 의료 영상 처리를 위한 기법을 제안한다. 제안 기법은 전반적인 조명 구성 요소를 무작위화하여 조명 조건의 변화를 모사함으로써, 영역화 모델이 보이지 않는 도메인에 대한 증강을 걸쳐 일반화를 진행한다. 두 번째 방법은 조명의 변화가 다양하고 복잡한 조명 패턴을 가지는 의료 영상 처리를 위한 기법으로 대장내시경 영상과 같은 이미지 처리에 적합하다. 해당 기법은 조명 구성 요소를 전역 및 지역 조명 구성 요소로 추가 분해하고 전역 조명 증강이 수행되어 보이지 않는 목표 도메인의 특성을 보다 효과적으로 포착하는 다양한 훈련 샘플을 합성하여 데이터를 증강한다.
이미지 공간에서 데이터 증강에 중점을 둔 제안된 조명 기반 방법과 달리, 특징 증강 기법은 무작위 컨볼루션 기반 적응형 스펙트럼 무작위 컨볼루션(ASRConv)을 진행하며 직접적으로 특징 공간에서 작동하여 고주파 아티팩트를 유발하지 않고 주파수 영역에서 원본 도메인 특징에 적절한 변형을 적용하고 모델이 조명 불변의 특징 표현을 학습한다. 제안 기법은 망막 및 대장내시경 이미지 데이터셋을 포함한 다양한 의료 이미지 영역화 데이터셋에서 수행한 다수의 실험에서 목표 도메인에서의 일반화 성능이 향상됨을 보여준다. 성능평가를 통해 데이터 증강 기반 방법들은 도메인 일반화를 통해 구역화의 정확도를 최대 9.60%까지 향상하였으며 특징 증강 기법은 기존 특징 증강 기반 도메인 일반화 기법 대비 3.07% 향상하였음을 보여준다.
결론적으로 본 논문이 기여한 바는 다음의 세가지로 요약할 수 있다. (1) 레티넥스 이론을 활용한 두 단계 조명 기반 데이터 증강 기법을 제안한다. (2) 도메인 일반화를 위한 특징 증강 방법인 ASRConv기법을 제안한다. (3) 다양한 의료 영상 데이터셋에서 다수의 성능평가를 통해 제안 기법의 효율성을 검증하였다. 본 논문의 기법을 적용하여 의료 영상에서의 도메인 변화에 대한 딥러닝 모델의 견고성을 향상시킴으로써 실제 임상 환경에서 자동 의료 영상 분석 시스템의 실용성을 개선하며 다양한 영상 촬영 환경에서 보다 정확하고 일관된 영상 분할 결과를 제공하여 의료진이 보다 정확한 의사 결정을 내릴 수 있도록 지원한다.

목차 (Table of Contents)

• 1 Introduction 1
• 1.1 Background 1
• 1.2 Problems 3
• 1.3 Contribution 4
• 1.4 Dissertation Outlines 5
• 2 Related Work 7
• 2.1 Medical Image Segmentation 7
• 2.2 Domain Adaptation 10
• 2.3 Domain Generalization 15
• 3 Data Augmentation-based Domain Generalization 23
• 3.1 Randomized Illumination Enhancement for Generalizable Segmentation 24
• 3.1.1 Preliminaries 25
• 3.1.2 Design Overview 26
• 3.1.3 TGCI: Metric for Retinex-based Image Decomposition Quality 30
• 3.2 Adversarial Global Illumination Augmentation for Generalizable Segmentation 32
• 3.2.1 Preliminaries 33
• 3.2.2 Design Overview 34
• 3.2.3 IViSen: Measuring Model Stability Under Varying Illumination 38
• 4 Feature Augmentation-based Domain Generalization 39
• 4.1 Adaptive Spectral Random Convolution for Generalizable Segmentation 40
• 4.1.1 Preliminaries 40
• 4.1.2 Design Overview 41
• 4.1.3 Objective Function 46
• 5 Experiments 47
• 5.1 Experimental Setup 47
• 5.2 Experiments on Data Augmentation-based DG Methods 49
• 5.2.1 Randomized Illumination Augmentation-based DG Method 49
• 5.2.2 Adversarial Global Illumination Augmentation-based DG Method 58
• 5.3 Experiments on Feature Augmentation-based DG Methods 68
• 5.3.1 Adaptive Spectral Random Convolution-based DG Method 68
• 5.4 Further Analysis 79
• 5.5 Conclusion 84
• 6 Conclusion 85
• References 89
• Appendix Nomenclature 101

참고문헌 (Reference)

1. Fast autoaugment, Sungwoong Kim, Ildoo Kim, Chiheon Kim and, Sungbin Lim, Taesup Kim, Advances in Neural Information Processing Systems, 32:6665–6675, , 2019

2. Wasserstein auto-encoders, Bernhard Schoelkopf, Olivier Bousquet, Ilya Tolstikhin, Sylvain Gelly and, arXiv preprint arXiv:1711.01558, , 2017

3. Generative adversarial nets, Ian Goodfellow, Bing Xu, David Warde- Farley, Sherjil Ozair, Jean Pouget-Abadie, Aaron Courville and, Mehdi Mirza, Yoshua Bengio, Advances in Neural Information Processing Systems, 27, , 2014

4. Colorectal cancer statistics, Lynn F Butterly, Kimberly D Miller, Andrea Cercek, Ann Goding Sauer, Ahmedin Jemal, Stacey A Fedewa, Joseph C Anderson, Robert A Smith and, Rebecca L Siegel, CA: A Cancer Journal for Clinicians, 70(3):145– 164, , 2020

5. Auto-encoding variational bayes, Max Welling, Diederik P Kingma and, arXiv preprint arXiv:1312.6114, , 2013

6. The retinex theory of color vision, Edwin H Land, 237(6):108– 129, , 1977

7. Beyond brightening low-light images, Wei Liu and, Xiaojie Guo, Yonghua Zhang, Jiayi Ma, Jiawan Zhang, 129(4):1013–1037, , 2021

8. Domain generalization with mixstyle, Yongxin Yang, Yu Qiao and, Tao Xiang, Kaiyang Zhou, arXiv preprint arXiv:2104.02008, , 2021

9. Kvasir-seg: A segmented polyp dataset, Håvard D Johansen, Pål Halvorsen, Dag Johansen and, Pia H Smedsrud, Debesh Jha, Thomas de Lange, Michael A Riegler, In International Conference on Multimedia Modeling, pages 451–462. Springer, , 2020

10. mixupBeyond empirical risk minimization, Hongyi Zhang, David Lopez-Paz, Yann N Dauphin and, Moustapha Cisse, arXiv preprint arXiv:1710.09412, , 2017

11. Metalearning in neural networks: A survey, Antreas Antoniou, Amos Storkey, Paul Micaelli and, Timothy Hospedales, 44(9):5149–5169, , 2021

12. Learning to learn single domain generalization, Long Zhao and, Fengchun Qiao, Xi Peng, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition pages 12556–12565, , 2020

13. Domain generalization by solving jigsaw puzzles, Antonio D’Innocente, Fabio M Carlucci, Barbara Caputo and, Tatiana Tommasi, Silvia Bucci, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2229– 2238, , 2019

14. Multi-scale retinex for color image enhancement, Daniel J Jobson and, Zia-ur Rahman, Glenn A Woodell, In Proceedings of 3rd IEEE International Conference on Image Processing, volume 3, pages 1003–1006. IEEE, , 1996

15. Image segmentation using deep learning: A survey, Demetri Terzopoulos, Yuri Boykov, Antonio Plaza, Shervin Minaee, Fatih Porikli, Nasser Kehtarnavaz and, 44(7):3523–3542, , 2021

16. Particleaugment: Samplingbased data augmentation, Alexander Tsaregorodtsev and, Vasileios Belagiannis, arXiv preprint arXiv:2106.08693, , 2021

17. Diagnostics and epidemiology of colorectal cancer, Frank T Kolligs, Visceral medicine, 32(3):158–164, , 2016

18. Unsupervised domain adaptation by backpropagation, Yaroslav Ganin and, Victor Lempitsky, In International Conference on Machine Learning, pages 1180–1189 PMLR, , 2015

19. Non-local u-nets for biomedical image segmentation, Shuiwang Ji, Na Zou, Dinggang Shen and, Zhengyang Wang, In volume 34, pages 6315–6322, , 2020

20. A fourier-based framework for domain generalization, Ya Zhang, Qi Tian, Ruipeng Zhang, Qinwei Xu, Yanfeng Wang and, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 14383–14392, , 2021

21. A joint intrinsic-extrinsic prior model for retinex, Bin Hu and, Kailing Guo, Dacheng Tao, Xianming Xu, Kui Jia, Bolun Cai, In Proceedings of the IEEE International Conference on Computer Vision, pages 4000–4009, , 2017

22. A review of domain adaptation without target labels, Marco Loog, WouterMKouw and, 43(3):766–785, , 2019

23. Deep retinex decomposition for low-light enhancement, WenjingWang, Wenhan Yang and, Jiaying Liu, ChenWei, arXiv preprint arXiv:1808.04560, , 2018

24. Mobilenetv2Inverted residuals and linear bottlenecks, Andrew Howard, Mark Sandler, Andrey Zhmoginov and, Menglong Zhu, Liang-Chieh Chen, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 4510–4520, , 2018

25. Source-free domain adaptation for image segmentation, Jose Dolz, Ismail Ben Ayed, Hoel Kervadec, Hervé Lombaert and, Mathilde Bateson, 82:102617, , 2022

26. Visualizing and understanding convolutional networks, Rob Fergus, Matthew D Zeiler and, In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland Proceedings, Part I 13, pages 818–833 Springer, , 2014

27. A survey on image data augmentation for deep learning, Taghi M Khoshgoftaar, Connor Shorten and, 6(1):1–48, , 2019

28. A survey on deep learning for skin lesion segmentation, Catarina Barata, Kumar Abhishek, Alceu Bissoto, Eduardo Valle, Sandra Avila, Ghassan Hamarneh, Zahra Mirikharaji, M Emre Celebi and, Medical Image Analysis, page 102863, 2023, , 2023

29. Domain adaptation for medical image analysis: a survey, Mingxia Liu, Hao Guan and, IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, , 2021

30. Fully convolutional networks for semantic segmentation, Trevor Darrell, Jonathan Long, Evan Shelhamer and, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3431–3440, , 2015

31. Vidit: Virtual image dataset for illumination transfer, Ruofan Zhou, Sabine Süsstrunk, Majed El Helou, Johan Barthas and, arXiv preprint arXiv:2005.05460, , 2020

32. A simple feature augmentation for domain generalization, Yanwei Fu and, Wei Li, Shaogang Gong, Pan Li, Da Li, Timothy M Hospedales, In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 8886–8895, , 2021

33. Autoaugment: Learning augmentation strategies from data, Ekin D Cubuk, Dandelion Mane, Barret Zoph, Vijay Vasudevan and, Quoc V Le, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 113–123, , 2019

34. Domain generalization with adversarial feature learning, Shiqi Wang and, Haoliang Li, Sinno Jialin Pan, Alex C Kot, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5400–5409, , 2018

35. Dynamic convolution: Attention over convolution kernels, Mengchen Liu, Dongdong Chen, Lu Yuan and, Xiyang Dai, Yinpeng Chen, Zicheng Liu, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11030–11039, , 2020

36. Attention u-net: Learning where to look for the pancreas, Jo Schlemper, Kensaku Mori, Ozan Oktay, Steven McDonagh, Mattias Heinrich, Kazunari Misawa, Bernhard Kainz et al, Loic Le Folgoc, Nils Y Hammerla, Matthew Lee, arXiv preprint arXiv:1804.03999, , 2018

37. Fda: Fourier domain adaptation for semantic segmentation, Yanchao Yang and, Stefano Soatto, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4085–4095, , 2020

38. Generalizing to unseen domains via distribution matching, Isabela Albuquerque, Mohammad Darvishi, Tiago H Falk and, Ioannis Mitliagkas, João Monteiro, arXiv preprint arXiv:1911.00804, , 2019

39. Nonlinear total variation based noise removal algorithms, Leonid I. Rudin, Stanley Osher and, Emad Fatemi, 60(1-4):259–268, , 1992

40. Understanding neural networks through deep visualization, Jason Yosinski, Jeff Clune, Anh Nguyen, Thomas Fuchs and, Hod Lipson, arXiv preprint arXiv:1506.06579, , 2015

41. Adaptive multiscale retinex for image contrast enhancement, Chang-Hsing Lee, Jau-Ling Shih, Chin-Chuan Han, Cheng-Chang Lien and, In 2013 International Conference on Signal-Image Technology Internet-Based Systems, pages 43–50. IEEE, , 2013

42. Feature-based style randomization for domain generalization, Yang Gao, Yinghuan Shi and, Yue Wang, Lei Qi, IEEE Transactions on Circuits and Systems for Video Technology, 32(8):5495–5509, , 2022

43. Kindling the darkness: A practical low-light image enhancer, Yonghua Zhang, Jiawan Zhang and, Xiaojie Guo, In Proceedings of the 27th ACM International Conference on Multimedia, pages 1632–1640, , 2019

44. Weighted res-unet for highquality retina vessel segmentation, Xiao Xiao, Zhiming Luo and, Shaozi Li, Shen Lian, In 9th International Conference on Information Technology in Medicine and Education (ITME), pages 327–331. IEEE, , 2018

45. Rethinking atrous convolution for semantic image segmentation, George Papandreou, Hartwig Adam, Liang-Chieh Chen, Florian Schroff and, arXiv preprint arXiv:1706.05587, , 2017

46. Style augmentation: data augmentation via style randomization, Boguslaw Obara, Toby P Breckon and, Amir Atapour Abarghouei, Philip TG Jackson, Stephen Bonner, In CVPR workshops, volume 6, pages 10–11, , 2019

47. U-netConvolutional networks for biomedical image segmentation, Thomas Brox, Philipp Fischer and, Olaf Ronneberger, In International Conference on Medical Image Computing and Computer-assisted Intervention, pages 234–241. Springer, , 2015

48. Mutualprototype adaptation for cross-domain polyp segmentation, Bulat Ibragimov and, Yixuan Yuan, Meilu Zhu, Xiaoqing Guo, Chen Yang, 25(10):3886–3897, , 2021

49. Multi-scale retinex improvement for nighttime image enhancement, Haoning Lin and, Zhenwei Shi, 125(24):7143–7148, , 2014

50. Towards automatic polyp detection with a polyp appearance model, Fernando Vilarino, Jorge Bernal, Javier Sánchez and, 45(9):3166–3182, , 2012

51. Generalizing to unseen domains via adversarial data augmentation, Riccardo Volpi, Ozan Sener, Vittorio Murino and, Silvio Savarese, Hongseok Namkoong, John C Duchi, In Proceedings of the 32nd International Conference on Neural Information Processing Systems, pages 5339—-5349, , 2018

52. Guidedmixup: an efficient mixup strategy guided by saliency maps, Minsoo Kang and, Suhyun Kim, volume 37, pages 1096–1104, 2023, , 2023

53. Progressive random convolutions for single domain generalization, Debasmit Das, Sungha Choi, Seunghan Yang, Sungrack Yun, Seokeon Choi, Hyunsin Park and, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10312–10322, 2023, , 2023

54. Lime: Low-light image enhancement via illumination map estimation, Haibin Ling, Xiaojie Guo, Yu Li and, 26(2):982–993, , 2016

55. Pranet: Parallel reverse attention network for polyp segmentation, Ge-Peng Ji, Deng-Ping Fan, Ling Shao, Huazhu Fu, Geng Chen, Jianbing Shen and, Tao Zhou, In International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), pages 263–273. Springer, , 2020

56. Rim-one: An open retinal image database for optic nerve evaluation, Francisco Fumero, José L Sanchez, Silvia Alayón, Jose Sigut and, M Gonzalez- Hernandez, In 2011 24th International Symposium on Computer-based Medical Systems (CBMS), pages 1–6. IEEE, , 2011

57. Fsdr: Frequency space domain randomization for domain generalization, Jiaxing Huang, Aoran Xiao and, Dayan Guan, Shijian Lu, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6891–6902, , 2021

58. Contrastdependent saturation adjustment for outdoor image enhancement, Mongi A Abidi and, Woon Cho, Joonki Paik, Jinbeum Jang, Shuhang Wang, 34(1):7–17, , 2017

59. Generalized lightness adaptation with channel selective normalization, Xin Jin, Zhiwei Xiong, Shenglong Zhou, Ruikang Xu, Man Zhou and, Jie Huang, Mingde Yao, In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 10668–10679, 2023, , 2023

60. Illumination decomposition for photograph with multiple light sources, Ling Zhang, Chunxia Xiao, Qingan Yan, Hua Zou and, Zheng Liu, 26(9):4114–4127, , 2017

61. Progressive domain expansion network for single domain generalization, Zhengze Yu, Yepeng Weng, Lei Li, Xiaoyue Mi, Xiaoya Li and, Juan Cao, Ziyao Huang, Ke Gao, Boyang Xia, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 224–233, , 2021

62. Domain generalization via model-agnostic learning of semantic features, Konstantinos Kamnitsas and, Ben Glocker, Daniel Coelho de Castro, Qi Dou, Advances in Neural Information Processing Systems, 32:6450–6461, , 2019

63. Illumination-based data augmentation for robust background subtraction, Dimitrios Sakkos, Edmond SL Ho, Hubert PH Shum and, In 2019 13th International Conference on Software, Knowledge, Information Management and Applications (SKIMA), pages 1–8. IEEE, , 2019

64. Missformer: an effective transformer for 2d medical image segmentation, Xiaohong Huang, Zhifang Deng, Ying Fu, Dandan Li, Xueguang Yuan and, IEEE transactions on medical imaging, , 2022

65. Understanding deep learning (still) requires rethinking generalization, Oriol Vinyals, Moritz Hardt, Samy Bengio, Chiyuan Zhang, Benjamin Recht and, 64(3):107–115, , 2021

66. 3d u-net: learning dense volumetric segmentation from sparse annotation, Thomas Brox and, Soeren S Lienkamp, Olaf Ronneberger, Ahmed Abdulkadir, Özgün Çiçek, In Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, Proceedings Part II 19 pages 424–432 Springer, , 2016

67. Uacanet: Uncertainty augmented context attention for polyp segmentation, Daijin Kim, Taehun Kim, Hyemin Lee and, In Proceedings of the 29th ACM International Conference on Multimedia, pages 2167–2175, , 2021

68. Disentangle domain features for cross-modality cardiac image segmentation, Chenhao Pei, Liqin Huang and, Fuping Wu, Xiahai Zhuang, 71:102078, , 2021

69. Efficient inference in fully connected crfs with gaussian edge potentials, Vladlen Koltun, Philipp Krähenbühl and, Advances in Neural Information Processing Systems, 24, , 2011

70. Faster autoaugment: Learning augmentation strategies using backpropagation, Hideki Nakayama, Ryuichiro Hataya, Kazuki Yoshizoe and, Jan Zdenek, In European Conference on Computer Vision, pages 1–16. Springer, , 2020

71. 1. medical color image enhancement: Problems, challenges recent techniques, Dibya Jyoti Bora, In Intelligent Multimedia Data Analysis, pages 1–18. De Gruyter, , 2019

72. Maxstyle: Adversarial style composition for robust medical image segmentation, Cheng Ouyang, Chen Chen, Matthew Sinclair, Zeju Li, Wenjia Bai and, Daniel Rueckert, In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 151–161. Springer, , 2022

73. Randaugment: Practical automated data augmentation with a reduced search space, Quoc V Le., Ekin D Cubuk, Barret Zoph, Jonathon Shlens and, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 702–703, , 2020

74. When unseen domain generalization is unnecessary? rethinking data augmentation, Thomas Sanford, Daguang Xu and, Andriy Myronenko, Dong Yang, Holger Roth, Xiaosong Wang, Stephanie Harmon, Ling Zhang, Baris Turkbey, Ziyue Xu, arXiv preprint arXiv:1906.03347, , 2019

75. Robust and generalizable visual representation learning via random convolutions, Colin Raffel and, Marc Niethammer, Zhenlin Xu, Deyi Liu, Junlin Yang, arXiv:2007.13003, , 2020

76. Segnet: A deep convolutional encoder-decoder architecture for image segmentation, Roberto Cipolla, Vijay Badrinarayanan, Alex Kendall and, 39(12):2481–2495, , 2017

77. Encoder-decoder with atrous separable convolution for semantic image segmentation, Liang-Chieh Chen, Hartwig Adam, George Papandreou, Florian Schroff and, Yukun Zhu, In Proceedings of the European Conference on Computer Vision (ECCV), pages 801–818, , 2018

78. Evolutionary approach for autoaugment using the thermodynamical genetic algorithm, Akira Terauchi and, Naoki Mori, In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 9851–9858, , 2021

79. Semantic image segmentation with deep convolutional nets and fully connected crfs, Iasonas Kokkinos, George Papandreou, Alan L Yuille, Kevin Murphy and, Liang-Chieh Chen, arXiv preprint arXiv:1412.7062, , 2014

80. Frequency-mixed single-source domain generalization for medical image segmentation, Huazhu Fu, Jiang Liu, Heng Li, Haojin Li, Yan Hu and, Xiuyun Su, Wei Zhao, In International Conference on Medical Image Computing and Computer- Assisted Intervention, pages 127–136. Springer, 2023, , 2023

81. Generalize ultrasound image segmentation via instant and plug play style transfer, Guangquan Zhou, Alejandro F Frangi, et al, Xiaoqiong Huang, Xin Yang, Rui Li, Rui Gao, Yankai Huang, Yi Xiong, Zhendong Liu, Yuanji Zhang, In 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI), pages 419–423. IEEE, , 2021

82. Self-attentive spatial adaptive normalization for cross-modality domain adaptation, Behzad Bozorgtabar and, Jean-Philippe Thiran, Manana Lortkipanidze, Guillaume Vray, Devavrat Tomar, 40(10):2926–2938, , 2021

83. Uretinex-net: Retinex-based deep unfolding network for low-light image enhancement, Wenhui Wu, Wenhan Yang and, Jian Weng, Xu Wang, Pingping Zhang, Jianmin Jiang, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5901–5910, , 2022

84. CutmixRegularization strategy to train strong classifiers with localizable features, Junsuk Choe and, Seong Joon Oh, Youngjoon Yoo, Dongyoon Han, Sangdoo Yun, Sanghyuk Chun, In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 6023–6032, , 2019

85. Dltta: Dynamic learning rate for test-time adaptation on crossdomain medical images, Qi Dou, Hongzheng Yang, Pheng Ann Heng and, Meirui Jiang, Quande Liu, Jianfeng Cao, Cheng Chen, IEEE Transactions on Medical Imaging, 41(12):3575–3586, , 2022

86. Illumination decomposition for material recoloring with consistent interreflections, Maneesh Agrawala, Robert Carroll, Ravi Ramamoorthi and, In ACM SIGGRAPH 2011 papers, pages 1–10, , 2011

87. Semantic data augmentation based distance metric learning for domain generalization, Zhibin Wang and, Hao Li, Qi Qian, Mengzhu Wang, Jianlong Yuan, In Proceedings of the 30th ACM International Conference on Multimedia, pages 3214–3223, , 2022

88. Aadg: automatic augmentation for domain generalization on retinal image segmentation, Yiqi Zhang, Yijin Huang, Li Lin, Junyan Lyu, Xiaoying Tang, Pujin Cheng and, IEEE Transactions on Medical Imaging, 41(12):3699–3711, , 2022

89. Tvconv: Efficient translation variant convolution for layout-aware visual processing, Jierun Chen, Tianlang He, S-H Gary Chan, Li Ma, Weipeng Zhuo, Sangtae Ha and, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12548–12558, , 2022

90. Causality-inspired single-source domain generalization for medical image segmentation, Daniel Rueckert, Cheng Ouyang, Zeju Li, Chen Qin, Wenjia Bai and, Surui Li, Chen Chen, 42(4):1095–1106, , 2022

91. Adversarial consistency for single domain generalization in medical image segmentation, Matthew Ragoza, Shaoan Xie, Yanwu Xu, Kayhan Batmanghelich, Mingming Gong and, Maxwell Reynolds, In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 671–681. Springer, , 2022

92. Endosim2real: Consistency learning-based domain adaptation for instrument segmentation, Anirban Mukhopadhyay and, Ronja Strömsdörfer, Manish Sahu, Stefan Zachow, In International Conference on Medical Image Computing and Computer-assisted Intervention, pages 784–794. Springer, , 2020

93. Source free domain adaptation for medical image segmentation with fourier style mining, Yixuan Yuan, Xiaoqing Guo, Zhen Chen and, Chen Yang, 79:102457, , 2022

94. A source-free domain adaptive polyp detection framework with style diversification flow, Yixuan Yuan, Xinyu Liu and, IEEE Transactions on Medical Imaging, 41(7):1897–1908, , 2022

95. Boundary constraint network with cross layer feature integration for polyp segmentation, Tianwei Zhou, Tianfu Wang, Guanghui Yue, Wanwan Han, Runmin Cong and, Bin Jiang, 26(8):4090–4099, , 2022

96. Cross-modal attention-guided convolutional network for multi-modal cardiac segmentation, Xinna Guo, Lei Wang and, Luping Zhou, Ming Yang, Ziqi Zhou, Yinghuan Shi, Wanqi Yang, In Machine Learning in Medical Imaging: 10th International Workshop, MLMI 2019, Held in Conjunction with MICCAI 2019, Shenzhen, China, Proceedings 10, pages 601–610. Springer, 2019., , 2019

97. Illumination transition image: parameter-based illumination estimation and re-rendering, Jianyi Liu, Shaoyi Du, Lei Xiong, Gaofeng Meng and, Nanning Zheng, In 2008 19th International Conference on Pattern Recognition, pages 1–4. IEEE, , 2008

98. Patch-based output space adversarial learning for joint optic disc and cup segmentation, Lequan Yu, Chi-Wing Fu and, Xin Yang, Pheng-Ann Heng, ShujunWang, IEEE Transactions on Medical Imaging, 38(11):2485–2495, , 2019

99. Structure-driven unsupervised domain adaptation for cross-modality cardiac segmentation, Guodong Wei, Nenglun Chen, Zhixu Du, Dinggang Shen and, Zhiming Cui, Changjian Li, Wenping Wang, Runnan Chen, Lei Yang, IEEE Transactions on Medical Imaging, 40(12):3604–3616, , 2021

100. Maximum-entropy adversarial data augmentation for improved generalization and robustness, Dimitris Metaxas, Xi Peng and, Ting Liu, Long Zhao, In Proceedings of the 34th International Conference on Neural Information Processing Systems, volume 33, pages 14435–14447, , 2020

101. Mixed color channels (mcc): A universal module for mixed sample data augmentation methods, Jianfeng Ma, Zhongyuan Jiang and, Bin Xiao, Yang Wei, In 2022 IEEE International Conference on Multimedia and Expo (ICME), pages 1–6. IEEE, , 2022

102. Unet++: Redesigning skip connections to exploit multiscale features in image segmentation, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh and, Zongwei Zhou, Jianming Liang, 39(6):1856–1867, , 2019

103. Toward embedded detection of polyps in wce images for early diagnosis of colorectal cancer, Juan Silva, Xavier Dray and, Aymeric Histace, Bertrand Granado, Olivier Romain, 9(2):283– 293, , 2014

104. H-denseunet: hybrid densely connected unet for liver and tumor segmentation from ct volumes, Chi-Wing Fu and, Pheng-Ann Heng, Xiaomeng Li, Xiaojuan Qi, Qi Dou, Hao Chen, 37(12):2663–2674, , 2018

105. Selective feature aggregation network with area-boundary constraints for polyp segmentation, Kai-Yu Tong, Yuqi Fang, Cheng Chen, Yixuan Yuan and, In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 302–310. Springer, , 2019

106. Domain randomization for transferring deep neural networks from simulation to the real world, Josh Tobin, Jonas Schneider, Rachel Fong, Alex Ray, Wojciech Zaremba and, Pieter Abbeel, In International Conference on Intelligent Robots and Systems (IROS), pages 23–30. IEEE, , 2017

107. Devil is in channels: Contrastive single domain generalization for medical image segmentation, Shishuai Hu, Yong Xia, Zehui Liao and, In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 14–23. Springer, 2023, , 2023

108. Domain generalization for medical imaging classification with linear-dependency regularization, YuFei Wang, Renjie Wan, Alex C Kot, Tie-Qiang Li and, Shiqi Wang, Haoliang Li, arXiv preprint arXiv:2009.12829, , 2020

109. Robust color medical image segmentation on unseen domain by randomized illumination enhancement, Byeong-Seok Shin, Yan Li and, Zuyu Zhang, 145:105427, , 2022

110. Rethinking data augmentation for single-source domain generalization in medical image segmentation, Zixian Su, Qiufeng Wang and, Jie Sun, Kaizhu Huang, Xi Yang, Kai Yao, volume 37, pages 2366–2374, 2023, , 2023

111. Embracing the disharmony in medical imaging: A simple and effective framework for domain adaptation, Christos Davatzikos, Pratik Chaudhari and, Rongguang Wang, 76:102309, , 2022