Automatic nerve tracking in confocal images of corneal subbasal epithelium

Abstract

The analysis of nerve structures in the cornea subbasal epithelium is of relevant clinical interest, as it provides information related to changes caused by surgical interventions, transplantation or diseases (i.e. diabetic neuropathy). We addressed the problem of recognizing and tracing corneal nerves in confocal microscopy images with a novel method based on a sparse tracking scheme. After a set of seed points is identified all over the image, nerves are traced by connecting seeds by means of minimum cost paths, whose weights have been estimated considering several directional measures. The performance of the method was assessed on a dataset of 30 images, from both normal and pathological subjects, whose nerves were traced manually to provide a ground truth reference. An average sensitivity of 0.85 and false detection rate of 0.05 were obtained, with an average running time of 25 seconds per image.