Asan Medical Center team uses generative AI that changes only skin tone to close skin-tone gaps in dermatology AI (report)

What happened

A research team led by Professor Kim Namkug at Asan Medical Center in Seoul developed a generative AI technique that changes only the skin color in photos of light-skinned patients to resemble darker skin while preserving lesion shape. The aim is to reduce the problem of skin-disease diagnostic AI accuracy varying by patients' skin color and race.

Provider claims

The team said the AI-generated photos are very close to real photos.

Why it matters

It is an approach that uses synthetic data to offset data bias in medical AI.

Sources

• Press 피부색 따라 성능 격차 보이던 '피부질환 진단 AI'... 서울아산병원 김남국 교수팀, ‘피부색 변환 기술’로 잡았다 AI News Korea