A Roughness Grading Method for Skin Surface Microstructure Using Deep Learning for the Assessment of Atopic Dermatitis

Abstract

In this paper, we propose a skin surface microstructure roughness grading method using deep learning for the assessment of atopic dermatitis. Since symptoms of atopic dermatitis are related to roughness of the skin microstructure (skin fold and skin ridge), we propose a method to classify roughness grades using deep learning. The proposed method can quantitatively provide useful information to assist clinical doctor in diagnosis even with a small amount of training data by proposing a new data augmentation method that takes skin roughness into account. We developed new classifiers for 11 grades and 6 grades of skin roughness, and obtained 87.1% accuracy in the case of 6 grades classification, which is similar to a clinical doctor's judgment.