Mitigating Individual Skin Tone Bias in Skin Lesion Classification through Distribution-Aware Reweighting

2025-12-09

AI Summary

Skin lesion classification models exhibit fairness bias due to continuous variation in skin tone, which traditional discrete subgroup-based fairness methods fail to capture adequately. Method: We propose a distribution-aware reweighting framework that models skin tone as a continuous variable, estimates its density via kernel density estimation (KDE), and introduces a distance-aware loss function that dynamically assigns sample weights based on local density and neighborhood distance in the tone distribution. Contribution/Results: This work is the first to formally define and evaluate individual-level fairness from a continuous distribution perspective, overcoming limitations of discrete subgroup partitioning that ignore intra-group heterogeneity. Extensive experiments employ 12 statistical divergence measures—including Wasserstein distance and Hellinger distance—along with Fidelity Similarity for distribution-aware weighting. The method consistently improves classification accuracy for underrepresented skin tones and enhances overall fairness across both CNN and Transformer architectures, outperforming existing reweighting baselines.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

Skin color has historically been a focal point of discrimination, yet fairness research in machine learning for medical imaging often relies on coarse subgroup categories, overlooking individual-level variations. Such group-based approaches risk obscuring biases faced by outliers within subgroups. This study introduces a distribution-based framework for evaluating and mitigating individual fairness in skin lesion classification. We treat skin tone as a continuous attribute rather than a categorical label, and employ kernel density estimation (KDE) to model its distribution. We further compare twelve statistical distance metrics to quantify disparities between skin tone distributions and propose a distance-based reweighting (DRW) loss function to correct underrepresentation in minority tones. Experiments across CNN and Transformer models demonstrate: (i) the limitations of categorical reweighting in capturing individual-level disparities, and (ii) the superior performance of distribution-based reweighting, particularly with Fidelity Similarity (FS), Wasserstein Distance (WD), Hellinger Metric (HM), and Harmonic Mean Similarity (HS). These findings establish a robust methodology for advancing fairness at individual level in dermatological AI systems, and highlight broader implications for sensitive continuous attributes in medical image analysis.

Problem

• Addresses individual fairness in skin lesion classification
• Mitigates skin tone bias using distribution-based reweighting
• Models skin tone as continuous attribute to correct underrepresentation

Innovation

• Treats skin tone as continuous attribute via kernel density estimation
• Proposes distance-based reweighting loss to correct minority underrepresentation
• Compares twelve statistical metrics to quantify distribution disparities