Skin Cancer Machine Learning Model Tone Bias

2024-10-08

arXiv.org

AI Summary

This paper identifies significant skin-tone bias in AI-based melanoma diagnosis models—even when trained on skin-tone-balanced datasets, models exhibit substantially lower malignant detection rates for darker-skinned individuals (disparate impact scores of 0.577 and 0.684, both below the fairness threshold of 0.80). Method: Using the ISIC dataset, we train a CNN-based binary classifier (benign vs. malignant) and conduct empirical analyses to isolate bias sources. Contribution/Results: We demonstrate—empirically and for the first time—that this disparity stems not from imbalanced training data distributions, but from intrinsic model mechanisms. This challenges the widely held assumption that dataset balancing alone ensures algorithmic fairness in clinical AI. Our findings provide novel insights into the origins of algorithmic bias in dermatological AI systems and motivate a paradigm shift toward fairness-aware model design, evaluation, and auditing frameworks in medical AI development.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy
• 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
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Background: Many open-source skin cancer image datasets are the result of clinical trials conducted in countries with lighter skin tones. Due to this tone imbalance, machine learning models derived from these datasets can perform well at detecting skin cancer for lighter skin tones. Any tone bias in these models could introduce fairness concerns and reduce public trust in the artificial intelligence health field. Methods: We examine a subset of images from the International Skin Imaging Collaboration (ISIC) archive that provide tone information. The subset has a significant tone imbalance. These imbalances could explain a model's tone bias. To address this, we train models using the imbalanced dataset and a balanced dataset to compare against. The datasets are used to train a deep convolutional neural network model to classify the images as malignant or benign. We then evaluate the models' disparate impact, based on selection rate, relative to dark or light skin tone. Results: Using the imbalanced dataset, we found that the model is significantly better at detecting malignant images in lighter tone resulting in a disparate impact of 0.577. Using the balanced dataset, we found that the model is also significantly better at detecting malignant images in lighter versus darker tones with a disparate impact of 0.684. Using the imbalanced or balanced dataset to train the model still results in a disparate impact well below the standard threshold of 0.80 which suggests the model is biased with respect to skin tone. Conclusion: The results show that typical skin cancer machine learning models can be tone biased. These results provide evidence that diagnosis or tone imbalance is not the cause of the bias. Other techniques will be necessary to identify and address the bias in these models, an area of future investigation.

Problem

• Skin cancer models show bias towards lighter skin tones.
• Imbalanced datasets contribute to tone bias in AI models.
• Current models fail to meet fairness standards for skin tone.

Innovation

• Used ISIC archive images with tone information
• Trained deep convolutional neural network models
• Evaluated models' disparate impact on skin tones