A measurement of epidermal thickness of fingertip skin from OCT images using convolutional neural network

Abstract

In this study, we proposed a method to measure the epidermal thickness (ET) of skin based on deep convolutional neural network, which was used to determine the boundaries of skin surface and the ridge portion in dermal–epidermis junction (DEJ) in cross-section optical coherence tomography (OCT) images of fingertip skin. The ET was calculated based on the row difference between the surface and the ridge top, which is determined by search the local maxima of boundary of the ridge portion. The results demonstrated that the region of ridge portion in DEJ was well determined and the ET measurement in this work can reduce the effect of the papillae valley in DEJ by 9.85%. It can be used for quantitative characterization of skin to differentiate the skin diseases.