Effect of Demographic Bias on Skin Lesion Classification
The study investigates the impact of demographic bias on skin lesion classification using ResNet-based models. It highlights that sex-specific training datasets can optimize model performance, while age biases favor younger groups. The research also emphasizes the need for targeted strategies to mitigate these biases in machine learning applications.
- ▪The study evaluates skin lesion classification performance focusing on demographic bias related to patient sex and age.
- ▪Sex-specific training datasets improved model performance, particularly for male patients in female-majority cases.
- ▪Reinforcing and adversarial learning schemes reduced bias gaps in balanced datasets but were less effective in male-majority settings.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2606.03214 (cs) [Submitted on 2 Jun 2026] Title:Effect of Demographic Bias on Skin Lesion Classification Authors:Ralf Raumanns, Gerard Schouten, Veronika Cheplygina, Josien P.W. Pluim View a PDF of the paper titled Effect of Demographic Bias on Skin Lesion Classification, by Ralf Raumanns and Gerard Schouten and Veronika Cheplygina and Josien P.W. Pluim View PDF HTML (experimental) Abstract:In this study, we evaluate the performance of skin lesion classification using ResNet-based convolutional models, focusing on the impact of demographic bias in training data, particularly variations in patient sex and age.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
