DDA: Dimensionality Driven Augmentation Search for Contrastive Learning in Laparoscopic Surgery

Yuning Zhou, Henry Badgery, Matthew Read, James Bailey, Catherine Davey
Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, PMLR 250:1898-1926, 2024.

Abstract

Self-supervised learning (SSL) has the potential for effective representation learning in medical imaging, but the choice of data augmentation is critical and domain-specific. It remains uncertain if general augmentation policies suit surgical applications. In this work, we automate the search for suitable augmentation policies through a new method called Dimensionality Driven Augmentation Search (DDA). DDA leverages the local dimensionality of deep representations as a proxy target, and differentiably searches for suitable data augmentation policies in contrastive learning. We demonstrate the effectiveness and efficiency of DDA in navigating a large search space and successfully identifying an appropriate data augmentation policy for laparoscopic surgery. We systematically evaluate DDA across three laparoscopic image classification and segmentation tasks, where it significantly improves over existing baselines. Furthermore, DDAś optimised set of augmentations provides insight into domain-specific dependencies when applying contrastive learning in medical applications. For example, while hue is an effective augmentation for natural images, it is not advantageous for laparoscopic images.

Cite this Paper

BibTeX
@InProceedings{pmlr-v250-zhou24b, title = {DDA: Dimensionality Driven Augmentation Search for Contrastive Learning in Laparoscopic Surgery}, author = {Zhou, Yuning and Badgery, Henry and Read, Matthew and Bailey, James and Davey, Catherine}, booktitle = {Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning}, pages = {1898--1926}, year = {2024}, editor = {Burgos, Ninon and Petitjean, Caroline and Vakalopoulou, Maria and Christodoulidis, Stergios and Coupe, Pierrick and Delingette, Hervé and Lartizien, Carole and Mateus, Diana}, volume = {250}, series = {Proceedings of Machine Learning Research}, month = {03--05 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v250/main/assets/zhou24b/zhou24b.pdf}, url = {https://proceedings.mlr.press/v250/zhou24b.html}, abstract = {Self-supervised learning (SSL) has the potential for effective representation learning in medical imaging, but the choice of data augmentation is critical and domain-specific. It remains uncertain if general augmentation policies suit surgical applications. In this work, we automate the search for suitable augmentation policies through a new method called Dimensionality Driven Augmentation Search (DDA). DDA leverages the local dimensionality of deep representations as a proxy target, and differentiably searches for suitable data augmentation policies in contrastive learning. We demonstrate the effectiveness and efficiency of DDA in navigating a large search space and successfully identifying an appropriate data augmentation policy for laparoscopic surgery. We systematically evaluate DDA across three laparoscopic image classification and segmentation tasks, where it significantly improves over existing baselines. Furthermore, DDAś optimised set of augmentations provides insight into domain-specific dependencies when applying contrastive learning in medical applications. For example, while hue is an effective augmentation for natural images, it is not advantageous for laparoscopic images.} }
Endnote
%0 Conference Paper %T DDA: Dimensionality Driven Augmentation Search for Contrastive Learning in Laparoscopic Surgery %A Yuning Zhou %A Henry Badgery %A Matthew Read %A James Bailey %A Catherine Davey %B Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2024 %E Ninon Burgos %E Caroline Petitjean %E Maria Vakalopoulou %E Stergios Christodoulidis %E Pierrick Coupe %E Hervé Delingette %E Carole Lartizien %E Diana Mateus %F pmlr-v250-zhou24b %I PMLR %P 1898--1926 %U https://proceedings.mlr.press/v250/zhou24b.html %V 250 %X Self-supervised learning (SSL) has the potential for effective representation learning in medical imaging, but the choice of data augmentation is critical and domain-specific. It remains uncertain if general augmentation policies suit surgical applications. In this work, we automate the search for suitable augmentation policies through a new method called Dimensionality Driven Augmentation Search (DDA). DDA leverages the local dimensionality of deep representations as a proxy target, and differentiably searches for suitable data augmentation policies in contrastive learning. We demonstrate the effectiveness and efficiency of DDA in navigating a large search space and successfully identifying an appropriate data augmentation policy for laparoscopic surgery. We systematically evaluate DDA across three laparoscopic image classification and segmentation tasks, where it significantly improves over existing baselines. Furthermore, DDAś optimised set of augmentations provides insight into domain-specific dependencies when applying contrastive learning in medical applications. For example, while hue is an effective augmentation for natural images, it is not advantageous for laparoscopic images.
APA
Zhou, Y., Badgery, H., Read, M., Bailey, J. & Davey, C.. (2024). DDA: Dimensionality Driven Augmentation Search for Contrastive Learning in Laparoscopic Surgery. Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 250:1898-1926 Available from https://proceedings.mlr.press/v250/zhou24b.html.

Related Material