Higher Chest X-ray Resolution Improves Classification Performance

TL;DR

This study demonstrates that training chest X-ray classification models at higher resolutions, like 1024 x 1024 pixels, significantly improves accuracy and pathology detection compared to lower resolutions, highlighting the importance of resolution in medical imaging.

Contribution

The paper provides empirical evidence that higher training resolutions enhance chest X-ray classification performance and pathology localization, challenging the common practice of using lower resolutions.

Findings

1024 x 1024 resolution yields highest AUC of 84.2%.

Lower resolutions like 256 x 256 hinder small pathology detection.

High resolution reduces reliance on spurious features.

Abstract

Deep learning models for image classification are often trained at a resolution of 224 x 224 pixels for historical and efficiency reasons. However, chest X-rays are acquired at a much higher resolution to display subtle pathologies. This study investigates the effect of training resolution on chest X-ray classification performance, using the chest X-ray 14 dataset. The results show that training with a higher image resolution, specifically 1024 x 1024 pixels, results in the best overall classification performance with a mean AUC of 84.2 % compared to 82.7 % when trained with 256 x 256 pixel images. Additionally, comparison of bounding boxes and GradCAM saliency maps suggest that low resolutions, such as 256 x 256 pixels, are insufficient for identifying small pathologies and force the model to use spurious discriminating features. Our code is publicly available at https://gitlab.lrz.de/IP/cxr-resolution