Who Does Your Algorithm Fail? Investigating Age and Ethnic Bias in the MAMA-MIA Dataset

TL;DR

This paper audits the fairness of breast cancer tumor segmentation models in the MAMA-MIA dataset, revealing age-related biases and the impact of data source aggregation on ethnic biases, highlighting fairness issues in medical image segmentation.

Contribution

It introduces a comprehensive fairness evaluation of segmentation models in medical imaging, focusing on age and ethnicity biases in the MAMA-MIA dataset, which is underexplored in this context.

Findings

Age bias persists against younger patients in segmentation.

Data source aggregation affects site-specific ethnic biases.

Biases may be linked to physiological factors affecting model performance.

Abstract

Deep learning models aim to improve diagnostic workflows, but fairness evaluation remains underexplored beyond classification, e.g., in image segmentation. Unaddressed segmentation bias can lead to disparities in the quality of care for certain populations, potentially compounded across clinical decision points and amplified through iterative model development. Here, we audit the fairness of the automated segmentation labels provided in the breast cancer tumor segmentation dataset MAMA-MIA. We evaluate automated segmentation quality across age, ethnicity, and data source. Our analysis reveals an intrinsic age-related bias against younger patients that continues to persist even after controlling for confounding factors, such as data source. We hypothesize that this bias may be linked to physiological factors, a known challenge for both radiologists and automated systems. Finally, we show how aggregating data from multiple data sources influences site-specific ethnic biases, underscoring the necessity of investigating data at a granular level.