Data Valuation for Medical Imaging Using Shapley Value: Application on A Large-scale Chest X-ray Dataset

TL;DR

This paper demonstrates that data Shapley can effectively identify low quality and mislabeled images in large-scale chest X-ray datasets, improving pneumonia detection models by filtering data based on valuation scores.

Contribution

The study applies data Shapley for data valuation in medical imaging, showing its effectiveness in identifying low quality and mislabeled data to enhance model performance.

Findings

Removing high Shapley value data decreases performance.

Removing low Shapley value data improves performance.

High Shapley data correlates with true pneumonia cases.

Abstract

The reliability of machine learning models can be compromised when trained on low quality data. Many large-scale medical imaging datasets contain low quality labels extracted from sources such as medical reports. Moreover, images within a dataset may have heterogeneous quality due to artifacts and biases arising from equipment or measurement errors. Therefore, algorithms that can automatically identify low quality data are highly desired. In this study, we used data Shapley, a data valuation metric, to quantify the value of training data to the performance of a pneumonia detection algorithm in a large chest X-ray dataset. We characterized the effectiveness of data Shapley in identifying low quality versus valuable data for pneumonia detection. We found that removing training data with high Shapley values decreased the pneumonia detection performance, whereas removing data with low Shapley values improved the model performance. Furthermore, there were more mislabeled examples in low Shapley value data and more true pneumonia cases in high Shapley value data. Our results suggest that low Shapley value indicates mislabeled or poor quality images, whereas high Shapley value indicates data that are valuable for pneumonia detection. Our method can serve as a framework for using data Shapley to denoise large-scale medical imaging datasets.