On the limits of cross-domain generalization in automated X-ray prediction

TL;DR

This study investigates the limits of cross-domain generalization in automated X-ray diagnosis, revealing that label shifts, not image shifts, primarily hinder model transferability across datasets.

Contribution

It provides a large-scale analysis of cross-domain generalization in X-ray prediction, highlighting the role of label shifts and discrepancies between model performance and agreement.

Findings

Label shifts, not image shifts, dominate generalization issues.

Models with similar performance can disagree significantly.

Regularization reveals variation in task similarity across datasets.

Abstract

This large scale study focuses on quantifying what X-rays diagnostic prediction tasks generalize well across multiple different datasets. We present evidence that the issue of generalization is not due to a shift in the images but instead a shift in the labels. We study the cross-domain performance, agreement between models, and model representations. We find interesting discrepancies between performance and agreement where models which both achieve good performance disagree in their predictions as well as models which agree yet achieve poor performance. We also test for concept similarity by regularizing a network to group tasks across multiple datasets together and observe variation across the tasks. All code is made available online and data is publicly available: https://github.com/mlmed/torchxrayvision