Automatic cell identification and analysis on in vivo reflectance confocal microscopy images of the human epidermis

Reflectance confocal microscopy (RCM) allows real-time in vivovisualization of the skin at cellular level. The study of RCM images provides information on the topological and geometrical properties of the epidermis. These may change in each layer of the epidermis, depending on the subject's age and the presence of certain dermatological conditions. Studying RCM images requires manual identification of cells to derive these properties which is time consuming and subject to human error, highlighting the need for an automated cell identification method. We propose an automated pipeline to analyze the structure of the skin in RCM images. The first step is to identify the region of interest (ROI) containing the epidermal cells. The second step is to identify individual cells in the segmented tissue area using an image filter. We then use prior biological knowledge to process the resulting detected cells, removing cells that are too small and re-applying the used filter locally on detected regions that are too big to be considered as a single cell. The results are evaluated both on simulated data and on manually annotated real RCM data. This study shows that automatic cellidentification can be achieved, with an accuracy (precision and recall) that matches the inter expert variability.