MoCo-CXR: MoCo Pretraining Improves Representation and Transferability of Chest X-ray Models

TL;DR

This paper introduces MoCo-CXR, a contrastive learning pretraining method for chest X-ray models that enhances their representation quality and transferability, especially with limited labeled data.

Contribution

MoCo-CXR adapts the Momentum Contrast method for medical imaging, improving chest X-ray model initialization and transferability across datasets and tasks.

Findings

Pretrained models outperform non-pretrained on pathology detection.

MoCo-CXR benefits are greatest with limited labeled data.

Transferability demonstrated on unseen tuberculosis dataset.

Abstract

Contrastive learning is a form of self-supervision that can leverage unlabeled data to produce pretrained models. While contrastive learning has demonstrated promising results on natural image classification tasks, its application to medical imaging tasks like chest X-ray interpretation has been limited. In this work, we propose MoCo-CXR, which is an adaptation of the contrastive learning method Momentum Contrast (MoCo), to produce models with better representations and initializations for the detection of pathologies in chest X-rays. In detecting pleural effusion, we find that linear models trained on MoCo-CXR-pretrained representations outperform those without MoCo-CXR-pretrained representations, indicating that MoCo-CXR-pretrained representations are of higher-quality. End-to-end fine-tuning experiments reveal that a model initialized via MoCo-CXR-pretraining outperforms its non-MoCo-CXR-pretrained counterpart. We find that MoCo-CXR-pretraining provides the most benefit with limited labeled training data. Finally, we demonstrate similar results on a target Tuberculosis dataset unseen during pretraining, indicating that MoCo-CXR-pretraining endows models with representations and transferability that can be applied across chest X-ray datasets and tasks.