Robust Fairness Vision-Language Learning for Medical Image Analysis

TL;DR

This paper proposes a framework to enhance fairness and robustness in medical vision-language models by adjusting training loss with dynamic data mining and distribution alignment, leading to improved equitable performance.

Contribution

It introduces a novel training framework combining Dynamic Bad Pair Mining and Sinkhorn distance to improve fairness and robustness in medical vision-language models.

Findings

Up to 8.6% improvement in equity-scaled AUC

Enhanced fairness across protected groups

Improved robustness in medical image analysis

Abstract

The advent of Vision-Language Models (VLMs) in medical image analysis has the potential to help process multimodal inputs and increase performance over traditional inference methods. However, when considering the domain in which these models will be implemented, fairness and robustness are important to ensure the model stays true for any patient. In this paper, we introduce a framework for ensuring robustness and fairness of VLM models. This framework modifies the loss function at training by identifying and adjusting faulty image-text pairs through a Dynamic Bad Pair Mining algorithm and also utilizing Sinkhorn distance to ensure the loss distributions of protected groups do not deviate from the total loss. Experimental testing of our framework shows up to a 8.6\% improvement when looking at equity-scaled AUC.