Pilot study of semiautomated localization of the dermal/epidermal junction in reflectance confocal microscopy images of skin

摘要

Reflectance confocal microscopy (RCM) continues to be translated toward the detection of skin cancersnin vivo. Automated image analysis may help clinicians and accelerate clinical acceptance of RCM. For screeningnand diagnosis of cancer, the dermal/epidermal junction (DEJ), at which melanomas and basal cell carcinomasnoriginate, is an important feature in skin. In RCM images, the DEJ is marked by optically subtle changes andnfeatures and is difficult to detect purely by visual examination. Challenges for automation of DEJ detection includenheterogeneity of skin tissue, high inter-, intra-subject variability, and low optical contrast. To cope with thesenchallenges, we propose a semiautomated hybrid sequence segmentation/classification algorithm that partitionsnz-stacks of tiles into homogeneous segments by fitting a model of skin layer dynamics and then classifies tilensegments as epidermis, dermis, or transitional DEJ region using texture features.We evaluate two different trainingnscenarios: 1. training and testing on portions of the same stack; 2. training on one labeled stack and testing onnone from a different subject with similar skin type. Initial results demonstrate the detectability of the DEJ in bothnscenarios with epidermis/dermis misclassification rates smaller than 10% and average distance from the expertnlabeled boundaries around 8.5 μm.