Papers
Topics
Authors
Recent
Search
2000 character limit reached

Learning to Generalize towards Unseen Domains via a Content-Aware Style Invariant Model for Disease Detection from Chest X-rays

Published 27 Feb 2023 in cs.CV, cs.AI, cs.LG, and eess.IV | (2302.13991v6)

Abstract: Performance degradation due to distribution discrepancy is a longstanding challenge in intelligent imaging, particularly for chest X-rays (CXRs). Recent studies have demonstrated that CNNs are biased toward styles (e.g., uninformative textures) rather than content (e.g., shape), in stark contrast to the human vision system. Radiologists tend to learn visual cues from CXRs and thus perform well across multiple domains. Motivated by this, we employ the novel on-the-fly style randomization modules at both image (SRM-IL) and feature (SRM-FL) levels to create rich style perturbed features while keeping the content intact for robust cross-domain performance. Previous methods simulate unseen domains by constructing new styles via interpolation or swapping styles from existing data, limiting them to available source domains during training. However, SRM-IL samples the style statistics from the possible value range of a CXR image instead of the training data to achieve more diversified augmentations. Moreover, we utilize pixel-wise learnable parameters in the SRM-FL compared to pre-defined channel-wise mean and standard deviations as style embeddings for capturing more representative style features. Additionally, we leverage consistency regularizations on global semantic features and predictive distributions from with and without style-perturbed versions of the same CXR to tweak the model's sensitivity toward content markers for accurate predictions. Our proposed method, trained on CheXpert and MIMIC-CXR datasets, achieves 77.32$\pm$0.35, 88.38$\pm$0.19, 82.63$\pm$0.13 AUCs(%) on the unseen domain test datasets, i.e., BRAX, VinDr-CXR, and NIH chest X-ray14, respectively, compared to 75.56$\pm$0.80, 87.57$\pm$0.46, 82.07$\pm$0.19 from state-of-the-art models on five-fold cross-validation with statistically significant results in thoracic disease classification.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (51)
  1. U. Kamal, M. Zunaed, N. B. Nizam, and T. Hasan, “Anatomy-XNet: An anatomy aware convolutional neural network for thoracic disease classification in chest X-rays,” IEEE J. Biomed. Health Inform., vol. 26, no. 11, pp. 5518–5528, 2022.
  2. Y. Zhang, L. Luo, Q. Dou, and P.-A. Heng, “Triplet attention and dual-pool contrastive learning for clinic-driven multi-label medical image classification,” Med Image Anal, vol. 86, p. 102772, 2023.
  3. H.-G. Jung, W.-J. Nam, H.-W. Kim, and S.-W. Lee, “Weakly supervised thoracic disease localization via disease masks,” Neurocomputing, vol. 517, pp. 34–43, 2023.
  4. Q. Li, Y. Lai, M. J. Adamu, L. Qu, J. Nie, and W. Nie, “Multi-level residual feature fusion network for thoracic disease classification in chest X-ray images,” IEEE Access, vol. 11, pp. 40 988–41 002, 2023.
  5. K. Zhou, Z. Liu, Y. Qiao, T. Xiang, and C. C. Loy, “Domain generalization: A survey,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 45, no. 4, p. 4396–4415, apr 2023.
  6. Y. Kang et al., “Improving domain generalization performance for medical image segmentation via random feature augmentation,” Methods, vol. 218, pp. 149–157, 2023.
  7. M. Islam, Z. Li, and B. Glocker, “Robustness stress testing in medical image classification,” in Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, 2023, pp. 167–176.
  8. L. Luo et al., “Deep mining external imperfect data for chest X-ray disease screening,” IEEE Trans. Med. Imaging, vol. 39, no. 11, pp. 3583–3594, 2020.
  9. R. Zhang, F. Yang, Y. Luo, J. Liu, and C. Wang, “Learning invariant representation for unsupervised domain adaptive thorax disease classification,” Pattern Recognit. Lett., vol. 160, pp. 155–162, 2022.
  10. Y. Feng et al., “Deep supervised domain adaptation for pneumonia diagnosis from chest X-ray images,” IEEE J. Biomed. Health Inform., vol. 26, no. 3, pp. 1080–1090, 2022.
  11. H. Wang and Y. Xia, “Domain-ensemble learning with cross-domain mixup for thoracic disease classification in unseen domains,” Biomed. Signal Process. Control, vol. 81, p. 104488, 2023.
  12. L. Zhang et al., “Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation,” IEEE Trans. Med. Imaging, vol. 39, no. 7, pp. 2531–2540, 2020.
  13. R. Yamashita, J. Long, S. Banda, J. Shen, and D. L. Rubin, “Learning domain-agnostic visual representation for computational pathology using medically-irrelevant style transfer augmentation,” IEEE Trans. Med. Imaging, vol. 40, no. 12, pp. 3945–3954, 2021.
  14. C. Li et al., “Domain generalization on medical imaging classification using episodic training with task augmentation,” Comput. Biol. Med., vol. 141, p. 105144, 2022.
  15. N. Baker, H. Lu, G. Erlikhman, and P. J. Kellman, “Deep convolutional networks do not classify based on global object shape,” PLoS Comput. Biol., vol. 14, no. 12, pp. 1–43, 12 2018.
  16. R. Geirhos, P. Rubisch, C. Michaelis, M. Bethge, F. A. Wichmann, and W. Brendel, “ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness,” in ICLR, 2019.
  17. P. T. Jackson, A. Atapour-Abarghouei, S. Bonner, T. P. Breckon, and B. Obara, “Style Augmentation: Data augmentation via style randomization,” in IEEE CVPR Workshops, 2019, pp. 83–92.
  18. K. Zhou, Y. Yang, Y. Qiao, and T. Xiang, “Domain generalization with mixstyle,” ICLR, 2021.
  19. Z. Zhong, Y. Zhao, G. H. Lee, and N. Sebe, “Adversarial style augmentation for domain generalized urban-scene segmentation,” Adv. neural inf. process. syst., vol. 35, pp. 338–350, 2022.
  20. H. Nam, H. Lee, J. Park, W. Yoon, and D. Yoo, “Reducing domain gap by reducing style bias,” in IEEE CVPR, 2021, pp. 8686–8695.
  21. Y. Wang, L. Qi, Y. Shi, and Y. Gao, “Feature-based style randomization for domain generalization,” IEEE Trans Circuits Syst Video Technol, vol. 32, no. 8, pp. 5495–5509, 2022.
  22. J. Irvin et al., “CheXpert: A large chest radiograph dataset with uncertainty labels and expert comparison,” AAAI, vol. 33, pp. 590–597, 7 2019.
  23. A. E. W. Johnson et al., “MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports,” Sci. Data, vol. 6, p. 317, 2019.
  24. E. P. Reis et al., “BRAX, Brazilian labeled chest X-ray dataset,” Sci. Data, vol. 9, p. 487, 2022.
  25. L. van der Maaten and G. Hinton, “Visualizing data using t-sne,” J. Mach. Learn. Res., vol. 9, no. 86, pp. 2579–2605, 2008.
  26. X. Huang and S. Belongie, “Arbitrary style transfer in real-time with adaptive instance normalization,” in IEEE ICCV, 2017, pp. 1510–1519.
  27. G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” in IEEE CVPR, 2017, pp. 2261–2269.
  28. O. Nuriel, S. Benaim, and L. Wolf, “Permuted AdaIN: Reducing the bias towards global statistics in image classification,” in IEEE CVPR, 2021, pp. 9477–9486.
  29. Z. Tang, Y. Gao, Y. Zhu, Z. Zhang, M. Li, and D. Metaxas, “CrossNorm and SelfNorm for generalization under distribution shifts,” in IEEE ICCV, 2021, pp. 52–61.
  30. X. Wang, Y. Peng, L. Lu, Z. Lu, M. Bagheri, and R. M. Summers, “ChestX-Ray8: Hospital-scale chest X-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases,” IEEE CVPR, pp. 3462–3471, 2017.
  31. H. Q. Nguyen et al., “VinDr-CXR: An open dataset of chest X-rays with radiologist’s annotations,” Sci. Data, vol. 9, p. 429, 2022.
  32. Y. Yang and S. Soatto, “FDA: Fourier domain adaptation for semantic segmentation,” in IEEE CVPR, 2020, pp. 4084–4094.
  33. H. Guan and M. Liu, “Domain adaptation for medical image analysis: A survey,” IEEE Trans. Biomed. Eng., vol. 69, no. 3, pp. 1173–1185, 2022.
  34. S. Khan, M. Asim, S. A. Chelloug, B. Abdelrahiem, S. Khan, and A. Musyafa, “A novel cluster matching-based improved kernel fisher criterion for image classification in unsupervised domain adaptation,” Symmetry, vol. 15, no. 6, 2023.
  35. K. Sanchez, C. Hinojosa, H. Arguello, D. Kouamé, O. Meyrignac, and A. Basarab, “CX-DaGAN: Domain adaptation for pneumonia diagnosis on a small chest X-ray dataset,” IEEE Trans. Med. Imaging, vol. 41, no. 11, pp. 3278–3288, 2022.
  36. P. Diao, A. Pai, C. Igel, and C. H. Krag, “Histogram-based unsupervised domain adaptation for medical image classification,” MICCAI, pp. 755–764, 2022.
  37. Y. Ganin et al., “Domain-adversarial training of neural networks,” J. Mach. Learn. Res., vol. 17, no. 1, p. 2096–2030, jan 2016.
  38. J. Wang et al., “Generalizing to unseen domains: A survey on domain generalization,” IEEE Trans Knowl Data Eng, vol. 35, no. 8, pp. 8052–8072, 2023.
  39. Y. Du et al., “Learning to learn with variational information bottleneck for domain generalization,” ECCV, pp. 200–216, 2020.
  40. M. Segu, A. Tonioni, and F. Tombari, “Batch normalization embeddings for deep domain generalization,” Pattern Recognit, vol. 135, p. 109115, 2023.
  41. S. Sankaranarayanan and Y. Balaji, “Chapter 6 - meta learning for domain generalization,” in Meta Learning With Medical Imaging and Health Informatics Applications, 2023, pp. 75–86.
  42. B. Garcia Santa Cruz, M. N. Bossa, J. Sölter, and A. D. Husch, “Public covid-19 X-ray datasets and their impact on model bias – a systematic review of a significant problem,” Med Image Anal, vol. 74, p. 102225, 2021.
  43. D. Ulyanov, A. Vedaldi, and V. Lempitsky, “Improved texture networks: Maximizing quality and diversity in feed-forward stylization and texture synthesis,” in IEEE CVPR, 2017, pp. 4105–4113.
  44. T. Miyato, S.-I. Maeda, M. Koyama, and S. Ishii, “Virtual adversarial training: A regularization method for supervised and semi-supervised learning,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 41, no. 8, pp. 1979–1993, 2019.
  45. J. Hu, L. Shen, and G. Sun, “Squeeze-and-Excitation networks,” in IEEE CVPR, 2018, pp. 7132–7141.
  46. L. A. Gatys, A. S. Ecker, and M. Bethge, “Image style transfer using convolutional neural networks,” in IEEE CVPR, 2016, pp. 2414–2423.
  47. X. Pan, P. Luo, J. Shi, and X. Tang, “Two at once: Enhancing learning and generalization capacities via IBN-Net,” in Eur. Conf. Comput. Vis., 2018, pp. 484–500.
  48. T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” in IEEE ICCV, 2017, pp. 2999–3007.
  49. B. T. Polyak and A. B. Juditsky, “Acceleration of stochastic approximation by averaging,” SIAM J Control Optim, vol. 30, no. 4, pp. 838–855, 1992.
  50. K. Sechidis, G. Tsoumakas, and I. Vlahavas, “On the stratification of multi-label data,” in Machine Learning and Knowledge Discovery in Databases, 2011, pp. 145–158.
  51. T. Karras, S. Laine, and T. Aila, “A style-based generator architecture for generative adversarial networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 43, no. 12, pp. 4217–4228, dec 2021.
Citations (4)
View on Semantic Scholar

Summary

No one has generated a summary of this paper yet.

Sign Up to Summarize

Paper to Video (Beta)

No one has generated a video about this paper yet.

Sign Up to Generate All Videos Subscribe on YouTube

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Sign Up to Generate

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Generate Now

Continue Learning

We haven't generated follow-up questions for this paper yet.

Generate Now

Collections

Sign up for free to add this paper to one or more collections.

Sign Up