Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces Label-free Hidden-Cohort Fairness (LHCF), a novel training paradigm for medical image analysis models that optimizes fairness across latent subpopulations discovered directly from image appearance using clustering. This approach addresses the limitations of traditional demographic-based fairness methods, which can overlook informative latent stratifications and suffer from data sparsity when considering multiple demographic attributes. Experiments on the proposed HIDFairBench benchmark demonstrate that LHCF achieves state-of-the-art fairness results across both single and multiple demographic attributes without using any demographic labels during training.
Forget optimizing for fairness across known demographics – clustering directly on image appearance reveals hidden cohorts that expose and mitigate deeper, previously unseen biases in medical imaging models.
Medical image analysis models can exhibit performance disparities across patient subgroups, threatening clinical safety and fairness. Existing methods typically address this issue by optimizing accuracy and fairness metrics for visible demographic attributes (e.g., sex or age) considered in isolation. This strategy not only overlooks potentially more informative latent stratifications, which may reveal deeper sources of model error and inequity, but also fails to scale when multiple demographic attributes are considered simultaneously due to the resulting sparsity of training data within each subgroup. We deal with these issues by introducing the label-free hidden-cohort fairness (LHCF) training paradigm that instead of maximizing fairness over visible demographic attributes, it optimizes fairness across latent subpopulations discovered from image appearance. By clustering images into K appearance-based cohorts and applying fairness optimization over them, LHCF uncovers underlying sources of model error and avoids the combinatorial sparsity of multi-demographic attributes, reducing disparities across both single and multiple demographic attributes. We demonstrate on our proposed fairness benchmark, HIDFairBench, that LHCF provides state-of-the-art fairness results on single and multiple demographic attributes, despite never using demographic labels for training. Our results position hidden-cohort fairness as a practical, scalable, and robust alternative to demographic-based fairness optimization for trustworthy medical image analysis.