Learning Unbiased Image Segmentation: A Case Study with Plain Knee Radiographs

TL;DR

This paper investigates gender and racial biases in deep learning-based knee bone segmentation from plain radiographs, revealing biases and proposing mitigation strategies to promote fairness and equitable healthcare outcomes.

Contribution

It uncovers visible biases in knee radiograph segmentation models and introduces mitigation strategies to ensure fairer, unbiased results in medical imaging.

Findings

Biases in gender and racial groups identified

Mitigation strategies reduce observed biases

Promotes equitable healthcare outcomes

Abstract

Automatic segmentation of knee bony anatomy is essential in orthopedics, and it has been around for several years in both pre-operative and post-operative settings. While deep learning algorithms have demonstrated exceptional performance in medical image analysis, the assessment of fairness and potential biases within these models remains limited. This study aims to revisit deep learning-powered knee-bony anatomy segmentation using plain radiographs to uncover visible gender and racial biases. The current contribution offers the potential to advance our understanding of biases, and it provides practical insights for researchers and practitioners in medical imaging. The proposed mitigation strategies mitigate gender and racial biases, ensuring fair and unbiased segmentation results. Furthermore, this work promotes equal access to accurate diagnoses and treatment outcomes for diverse patient populations, fostering equitable and inclusive healthcare provision.