Korean J Dermatol.  2018 Aug;56(7):421-425.

Acne Severity Scoring Using Deep Learning

• Lee SJ
• Jeon JS
• Ryu JY
• Song HJ
• Jo YJ
• Bang CH
• Park YM
• Lee JH

Abstract

BACKGROUND
Acne is a chronic inflammatory disease of the pilosebaceous unit, mainly on the face. It can have various clinical manifestations and should be appropriately treated based on the severity. In Korea, the "˜Korea Acne Severity Rating System (KAGS)' is a standardized index to determine the severity of acne according to specific Korean characteristics. However, the actual use of the KAGS in clinical settings has been limited.
OBJECTIVE
We sought to analyze whether we could effectively measure acne severity using a deep learning algorithm, which is an image learning method.
METHODS
Acne severity was classified into three levels of mild, moderate, and severe based on the KAGS, and learning and verification were performed using the CNN (Convolutional Neural Network), a deep learning technique.
RESULTS
GoogLeNet's Inception-v3 algorithm showed the highest accuracy at 86.7%.
CONCLUSION
This study confirmed that the use of a deep learning algorithm may facilitate the scoring of acne severity.

Keyword

Acne severity; Convolutional Neural Network Korean acne grading system; Deep learning

MeSH Terms

Acne Vulgaris*
Korea
Learning*
Methods