Estimation of skin surface roughness in vivo based on optical coherence tomography combined with convolutional neural network

The texture of human skin is influenced by both external and internal factors, and changes in wrinkles can most directly reflect the state of the skin. Skin roughness is primarily used to quantify the wrinkle features of the skin. Therefore, effective and accurate quantification of skin roughness is essential in skincare, medical treatment, and product development. This study proposes a method for estimating the skin surface roughness using optical coherence tomography (OCT) combined with a convolutional neural network (CNN). The proposed algorithm is validated through a roughness standard plate. Then, the experimental results revealed that skin surface roughness including arithmetic mean roughness and depth of roughness depends on age and gender. The advantage of the proposed method based on OCT is that it can reduce the effect of the skin surface's natural curvature on roughness. In addition, the method is combined with the epidermal thickness and dermal attenuation coefficient for multi-parameter characterization of skin features. It could be seen as a potential tool for understanding the aging process and developing strategies to maintain and enhance skin health and appearance.