Movatterモバイル変換

[0]ホーム
URL:

Skip to main content

Advertisement

Springer Nature Link
Log in

Addressing Catastrophic Forgetting by Modulating Global Batch Normalization Statistics for Medical Domain Expansion

  • Conference paper
  • First Online:

Abstract

Model brittleness across datasets is a key concern when deploying deep learning models in real-world medical settings. One approach is to fine-tune the model on subsequent datasets after training on the original dataset. However, this degrades model performance on the original dataset, a phenomenon known ascatastrophic forgetting. We develop an approach to address catastrophic forgetting by combining elastic weight consolidation with a simple yet novel modulation of global batch normalization statistics under two scenarios: expanding the domain across 1) imaging systems and 2) hospital institutions. Focusing on the clinical use case of mammographic breast density detection, we show that our approach empirically outperforms several other state-of-the-art approaches and provides theoretical justification for the efficacy of batch normalization modulation, demonstrating the potential of our approach to deploying clinical deep learning models requiring domain expansion.

S. Gupta and K. Chang—Co-first authors.

J. Kalpathy-Cramer and P. Singh—Co-senior authors.

This is a preview of subscription content,log in via an institution to check access.

Access this chapter

Subscribe and save

Springer+ Basic
¥17,985 /Month
  • Get 10 units per month
  • Download Article/Chapter or eBook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

Chapter
JPY 3498
Price includes VAT (Japan)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
JPY 5719
Price includes VAT (Japan)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
JPY 7149
Price includes VAT (Japan)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide -see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Similar content being viewed by others

Notes

  1. 1.

    Code Availability: All code from our method is available at:https://github.com/QTIM-Lab/MedicalDomainExpansion.

  2. 2.

    Although we were able to replicate their results (perform domain expansion with minimal CF) using a two-layered MLP (with dropout), we were unable to achieve high performance using a Resnet50 architecture. One possible reason could be that MLP, due to its fully connected layers, is somewhat blind to the permutations and hence does not forget much from task 1 when trained on task 2.

  3. 3.

    For a dropout probability of 0.10, accuracies for task 1 and 2 were 0.36 and 0.86 respectively. At higher dropout probabilities, the model was unable to converge for task 2.

References

  1. Chang, K., et al.: Multi-institutional assessment and crowdsourcing evaluation of deep learning for automated classification of breast density. J. Am. Coll. Radiol.17(12), 1653–1662 (2020)

    Google Scholar 

  2. Chen, T., Kornblith, S., Norouzi, M., Hinton, G.: A simple framework for contrastive learning of visual representations. arXiv preprintarXiv:2002.05709 (2020)

  3. Esteva, A., Kuprel, B., Novoa, R.A., Ko, J., Swetter, S.M., Blau, H.M., Thrun, S.: Dermatologist-level classification of skin cancer with deep neural networks. Nature542(7639), 115–118 (2017)

    Article  Google Scholar 

  4. Goodfellow, I.J., Mirza, M., Xiao, D., Courville, A., Bengio, Y.: An empirical investigation of catastrophic forgetting in gradient-based neural networks. arXiv preprintarXiv:1312.6211 (2013)

  5. He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016)

    Google Scholar 

  6. Ioffe, S., Szegedy, C.: Batch normalization: accelerating deep network training by reducing internal covariate shift. arXiv preprintarXiv:1502.03167 (2015)

  7. Karani, N., Chaitanya, K., Baumgartner, C., Konukoglu, E.: A lifelong learning approach to brain MR segmentation across scanners and protocols. In: International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 476–484. Springer (2018).https://doi.org/10.1007/978-3-030-00928-1_54

  8. Keavey, E., Phelan, N., O’Connell, A., Flanagan, F., O’Doherty, A., Larke, A., Connors, A.: Comparison of the clinical performance of three digital mammography systems in a breast cancer screening programme. Br. J. Radiol.85(1016), 1123–1127 (2012)

    Article  Google Scholar 

  9. Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. arXiv preprintarXiv:1412.6980 (2014)

  10. Kirkpatrick, J., Pascanu, R., Rabinowitz, N., Veness, J., Desjardins, G., Rusu, A.A., Milan, K., Quan, J., Ramalho, T., Grabska-Barwinska, A., et al.: Overcoming catastrophic forgetting in neural networks. Proc. Natl. Acad. Sci.114(13), 3521–3526 (2017)

    Article MathSciNet  Google Scholar 

  11. Li, M.D., et al.: Siamese neural networks for continuous disease severity evaluation and change detection in medical imaging. NPJ digital medicine3(1), 1–9 (2020)

    Article MathSciNet  Google Scholar 

  12. Li, Y., Wang, N., Shi, J., Liu, J., Hou, X.: Revisiting batch normalization for practical domain adaptation. arXiv preprintarXiv:1603.04779 (2016)

  13. Liberman, L., Abramson, A.F., Squires, F.B., Glassman, J., Morris, E., Dershaw, D.D.: The breast imaging reporting and data system: positive predictive value of mammographic features and final assessment categories. AJR Am. J. Roentgenol.171(1), 35–40 (1998)

    Article  Google Scholar 

  14. Ly, A., Marsman, M., Verhagen, J., Grasman, R., Wagenmakers, E.J.: A tutorial on fisher information (2017)

    Google Scholar 

  15. McInnes, L., Healy, J., Melville, J.: UMAP: uniform manifold approximation and projection for dimension reduction. arXiv preprintarXiv:1802.03426 (2018)

  16. Mirzadeh, S.I., Farajtabar, M., Ghasemzadeh, H.: Dropout as an implicit gating mechanism for continual learning. arXiv preprintarXiv:2004.11545 (2020)

  17. Mohamed, A., Berg, W., Peng, H., Luo, Y., Jankowitz, R., Wuand, S.: A deep learning method for classifying mammographic breast density categories. Med. Phys.45(1), 314–321 (2017)

    Article  Google Scholar 

  18. Pisano, E.D., Gatsonis, C., Hendrick, E., Yaffe, M., Baum, J.K., Acharyya, S., Conant, E.F., Fajardo, L.L., Bassett, L., D’Orsi, C., et al.: Diagnostic performance of digital versus film mammography for breast-cancer screening. N. Engl. J. Med.353(17), 1773–1783 (2005)

    Article  Google Scholar 

  19. Razzaghi, H., Troester, M.A., Gierach, G.L., Olshan, A.F., Yankaskas, B.C., Millikan, R.C.: Mammographic density and breast cancer risk in white and african american women. Breast Cancer Res. Treat.135(2), 571–580 (2012)

    Article  Google Scholar 

  20. Richard, L., Gary, K.: The measurement of observer agreement for categorical data. arxiv e-prints, page. Biometrics33, 159–174 (1977)

    Google Scholar 

  21. Roth, H.R., et al.: Federated learning for breast density classification: a real-world implementation. LNCS, pp. 181–191 (2020).https://doi.org/10.1007/978-3-030-60548-3_18,http://dx.doi.org/10.1007/978-3-030-60548-3_18

  22. Sheller, M.J., Edwards, B., Reina, G.A., Martin, J., Pati, S., Kotrotsou, A., Milchenko, M., Xu, W., Marcus, D., Colen, R.R., et al.: Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data. Sci. Rep.10(1), 1–12 (2020)

    Article  Google Scholar 

  23. Sprague, B.L., Conant, E.F., Onega, T., Garcia, M.P., Beaber, E.F., Herschorn, S.D., Lehman, C.D., Tosteson, A.N., Lacson, R., Schnall, M.D., et al.: Variation in mammographic breast density assessments among radiologists in clinical practice: a multicenter observational study. Ann. Intern. Med.165(7), 457–464 (2016)

    Article  Google Scholar 

  24. Sprague, B.L., et al.: Variation in mammographic breast density assessments among radiologists in clinical practice: a multicenter observational study. Ann. Intern. Med.165(7), 457–464 (2016)

    Article  Google Scholar 

  25. Yu, T., Bagdasaryan, E., Shmatikov, V.: Salvaging federated learning by local adaptation. arXiv preprintarXiv:2002.04758 (2020)

  26. Zagoruyko, S., Komodakis, N.: Wide residual networks. arXiv preprintarXiv:1605.07146 (2016)

  27. Zech, J.R., Badgeley, M.A., Liu, M., Costa, A.B., Titano, J.J., Oermann, E.K.: Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: a cross-sectional study. PLoS Med.15(11), e1002683 (2018)

    Google Scholar 

  28. Zeng, G., Chen, Y., Cui, B., Yu, S.: Continual learning of context-dependent processing in neural networks. Nature Machine Intelligence1(8), 364–372 (2019)

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Massachusetts General Hospital, Boston, MA, USA

    Sharut Gupta, Mehak Aggarwal, Mishka Gidwani, Jay Patel & Christopher P. Bridge

  2. Massachusetts Institute of Technology, Cambridge, MA, USA

    Sharut Gupta, Aakanksha Rana, Vibha Agarwal & Charles Lu

  3. Stanford University, Palo Alto, CA, USA

    Ken Chang & Daniel L. Rubin

  4. The University of Hong Kong, Pok Fu Lam, Hong Kong

    Liangqiong Qu

  5. Harvard Medical School, Boston, MA, USA

    Syed Rakin Ahmed & Katharina Hoebel

  6. Carnegie-Mellon University, Pittsburgh, PA, USA

    Nishanth Arun & Ashwin Vaswani

  7. The University of Texas at Austin, Austin, TX, USA

    Shruti Raghavan

  8. University of Colorado School of Medicine, Aurora, CO, USA

    Jayashree Kalpathy-Cramer & Praveer Singh

Authors
  1. Sharut Gupta

    You can also search for this author inPubMed Google Scholar

  2. Ken Chang

    You can also search for this author inPubMed Google Scholar

  3. Liangqiong Qu

    You can also search for this author inPubMed Google Scholar

  4. Aakanksha Rana

    You can also search for this author inPubMed Google Scholar

  5. Syed Rakin Ahmed

    You can also search for this author inPubMed Google Scholar

  6. Mehak Aggarwal

    You can also search for this author inPubMed Google Scholar

  7. Nishanth Arun

    You can also search for this author inPubMed Google Scholar

  8. Ashwin Vaswani

    You can also search for this author inPubMed Google Scholar

  9. Shruti Raghavan

    You can also search for this author inPubMed Google Scholar

  10. Vibha Agarwal

    You can also search for this author inPubMed Google Scholar

  11. Mishka Gidwani

    You can also search for this author inPubMed Google Scholar

  12. Katharina Hoebel

    You can also search for this author inPubMed Google Scholar

  13. Jay Patel

    You can also search for this author inPubMed Google Scholar

  14. Charles Lu

    You can also search for this author inPubMed Google Scholar

  15. Christopher P. Bridge

    You can also search for this author inPubMed Google Scholar

  16. Daniel L. Rubin

    You can also search for this author inPubMed Google Scholar

  17. Jayashree Kalpathy-Cramer

    You can also search for this author inPubMed Google Scholar

  18. Praveer Singh

    You can also search for this author inPubMed Google Scholar

Corresponding author

Correspondence toPraveer Singh.

Editor information

Editors and Affiliations

  1. University of Catania, Catania, Italy

    Federica Proietto Salanitri

  2. University of KwaZulu-Natal, Durban, South Africa

    Serestina Viriri

  3. Northwestern University, Chicago, IL, USA

    Ulaş Bağcı

  4. University of Wisconsin-Madison, Madison, WI, USA

    Pallavi Tiwari

  5. Boston University, Boston, MA, USA

    Boqing Gong

  6. University of Catania, Catania, Italy

    Concetto Spampinato

  7. University of Catania, Catania, Italy

    Simone Palazzo

  8. University of Catania, Catania, Italy

    Giovanni Bellitto

  9. National Technical University of Athens, Zografou, Greece

    Nancy Zlatintsi

  10. National Technical University of Athens, Zografou, Greece

    Panagiotis Filntisis

  11. University of Washington, Seattle, WA, USA

    Cecilia S. Lee

  12. University of Washington, Seattle, WA, USA

    Aaron Y. Lee

Rights and permissions

Copyright information

© 2025 The Author(s), under exclusive license to Springer Nature Switzerland AG

About this paper

Check for updates. Verify currency and authenticity via CrossMark

Cite this paper

Gupta, S.et al. (2025). Addressing Catastrophic Forgetting by Modulating Global Batch Normalization Statistics for Medical Domain Expansion. In: Proietto Salanitri, F.,et al. Artificial Intelligence in Pancreatic Disease Detection and Diagnosis, and Personalized Incremental Learning in Medicine. PILM AIPAD 2024 2024. Lecture Notes in Computer Science, vol 15197. Springer, Cham. https://doi.org/10.1007/978-3-031-73483-0_6

Download citation

Publish with us

Societies and partnerships

Access this chapter

Subscribe and save

Springer+ Basic
¥17,985 /Month
  • Get 10 units per month
  • Download Article/Chapter or eBook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

Chapter
JPY 3498
Price includes VAT (Japan)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
JPY 5719
Price includes VAT (Japan)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
JPY 7149
Price includes VAT (Japan)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide -see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

[8]ページ先頭
©2009-2025 Movatter.jp