A novel Dermo-epidermal localization algorithm for Swept source OCT images of human skin

摘要

Optical coherence tomography (OCT) is a noninvasive diagnostic method that offers a view into the superficial layers of the skin in vivo in real-time. OCT delivers morphological images of microstructures within the skin. Epidermal thickness in OCT images is of paramount importance, since dermo-epidermal junction (DEJ) location alteration is the start of several skin abnormalities. Due to the presence of speckle noise, devising an algorithm for locating DEJ in the OCT images is challenging. In this study we propose a semi-automatic DEJ detection algorithm based on graph theory that is resistant to speckle. In this novel approach we use attenuation map as a complementary feature compared to the previous methods that are mainly based on the intensity information. The method is based on converting border segmentation problem to the shortest path problem using graph theory. To smooth borders, we introduced a thinning fuzzy system enabling closer match to manual segmentation. Subsequently, an averaged A-scan analysis is performed to obtain the mean epidermal thickness. The DEJ detection method is performed on 96 B-Scan OCT skin images taken from different sites of body of healthy individuals. The results are evaluated based on several expert's visual analysis.