Enhancing Fairness in Skin Lesion Classification for Medical Diagnosis Using Prune Learning
Kuniko Paxton, Koorosh Aslansefat, Dhavalkumar Thakker, Yiannis Papadopoulos, Tanaya Maslekar · 2025
Read Enhancing Fairness in Skin Lesion Classification for Medical Diagnosis Using Prune Learning by Kuniko Paxton, Koorosh Aslansefat, Dhavalkumar Thakker, Yiannis Papadopoulos, Tanaya Maslekar (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-00745.txt.
Description
Kuniko Paxton leads five authors in total, with Koorosh Aslansefat, Dhavalkumar Thakker, Yiannis Papadopoulos, and Tanaya Maslekar, on fairness in skin lesion classification. Deep learning aids medical diagnosis, but models can be biased by skin color, and verifying fairness is hard because skin tones are tough to classify and checks are costly.
Paxton and her co-authors calculate the skewness, the asymmetry, of feature maps in a VGG network and of patches and attention heads in a Vision Transformer. Guided by those measures, their prune learning method cuts channels tied to skin tone and focuses on the lesion area. The approach lowers compute cost and model size while mitigating bias without conventional statistical methods.
Catalog record
| CREATOR | Kuniko Paxton, Koorosh Aslansefat, Dhavalkumar Thakker, Yiannis Papadopoulos, Tanaya Maslekar |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-00745