Exploring Image Augmentations for Siamese Representation Learning with Chest X-Rays

TL;DR

This paper systematically evaluates the impact of various image augmentations on Siamese network representations for chest X-ray analysis, demonstrating that certain augmentations significantly improve robustness and out-of-distribution generalization in medical imaging.

Contribution

It provides a comprehensive assessment of augmentation strategies for Siamese learning in medical images, identifying effective augmentations that enhance robustness and transferability.

Findings

Certain augmentations improve out-of-distribution generalization by up to 20%.

Robust representations are achieved with specific augmentation sets.

Zero-shot transfer and linear probes outperform supervised baselines.

Abstract

Image augmentations are quintessential for effective visual representation learning across self-supervised learning techniques. While augmentation strategies for natural imaging have been studied extensively, medical images are vastly different from their natural counterparts. Thus, it is unknown whether common augmentation strategies employed in Siamese representation learning generalize to medical images and to what extent. To address this challenge, in this study, we systematically assess the effect of various augmentations on the quality and robustness of the learned representations. We train and evaluate Siamese Networks for abnormality detection on chest X-Rays across three large datasets (MIMIC-CXR, CheXpert and VinDR-CXR). We investigate the efficacy of the learned representations through experiments involving linear probing, fine-tuning, zero-shot transfer, and data efficiency. Finally, we identify a set of augmentations that yield robust representations that generalize well to both out-of-distribution data and diseases, while outperforming supervised baselines using just zero-shot transfer and linear probes by up to 20%. Our code is available at https://github.com/StanfordMIMI/siaug.