FairSkin: Fair Diffusion for Skin Disease Image Generation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhang, Ruichen, Yao, Yuguang, Tan, Zhen, Li, Zhiming, Wang, Pan, Liu, Huan, Hu, Jingtong, Liu, Sijia, Chen, Tianlong
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915000453431296
author Zhang, Ruichen
Yao, Yuguang
Tan, Zhen
Li, Zhiming
Wang, Pan
Liu, Huan
Hu, Jingtong
Liu, Sijia
Chen, Tianlong
author_facet Zhang, Ruichen
Yao, Yuguang
Tan, Zhen
Li, Zhiming
Wang, Pan
Liu, Huan
Hu, Jingtong
Liu, Sijia
Chen, Tianlong
contents Image generation is a prevailing technique for clinical data augmentation for advancing diagnostic accuracy and reducing healthcare disparities. Diffusion Model (DM) has become a leading method in generating synthetic medical images, but it suffers from a critical twofold bias: (1) The quality of images generated for Caucasian individuals is significantly higher, as measured by the Frechet Inception Distance (FID). (2) The ability of the downstream-task learner to learn critical features from disease images varies across different skin tones. These biases pose significant risks, particularly in skin disease detection, where underrepresentation of certain skin tones can lead to misdiagnosis or neglect of specific conditions. To address these challenges, we propose FairSkin, a novel DM framework that mitigates these biases through a three-level resampling mechanism, ensuring fairer representation across racial and disease categories. Our approach significantly improves the diversity and quality of generated images, contributing to more equitable skin disease detection in clinical settings.
format Preprint
id arxiv_https___arxiv_org_abs_2410_22551
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle FairSkin: Fair Diffusion for Skin Disease Image Generation
Zhang, Ruichen
Yao, Yuguang
Tan, Zhen
Li, Zhiming
Wang, Pan
Liu, Huan
Hu, Jingtong
Liu, Sijia
Chen, Tianlong
Computer Vision and Pattern Recognition
Image generation is a prevailing technique for clinical data augmentation for advancing diagnostic accuracy and reducing healthcare disparities. Diffusion Model (DM) has become a leading method in generating synthetic medical images, but it suffers from a critical twofold bias: (1) The quality of images generated for Caucasian individuals is significantly higher, as measured by the Frechet Inception Distance (FID). (2) The ability of the downstream-task learner to learn critical features from disease images varies across different skin tones. These biases pose significant risks, particularly in skin disease detection, where underrepresentation of certain skin tones can lead to misdiagnosis or neglect of specific conditions. To address these challenges, we propose FairSkin, a novel DM framework that mitigates these biases through a three-level resampling mechanism, ensuring fairer representation across racial and disease categories. Our approach significantly improves the diversity and quality of generated images, contributing to more equitable skin disease detection in clinical settings.
title FairSkin: Fair Diffusion for Skin Disease Image Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.22551