Segmentation of nailfold capillaries from microscopy video sequences

Abstract

Nailfold capillary microscopy examination has been used for long time as a non invasive technique for the diagnosis of connective tissue diseases. Computer systems helping the physician during the examination have been developed, but they still require a certain level of manual intervention. A first problem that need yet to be solved is the segmentation of the capillaries in the video sequence. In this paper we propose a software pipeline for the segmentation of the capillaries from video sequences of the nailfold. The core of the segmentation proposed is the combination into a single segmentation map of the output of a battery of simple segmentation algorithms, based only on the grey levels. The system has been tested on a set of data, for which a ground truth segmentation was obtained manually, reporting satisfactory results. We advocate that the inclusion in the battery of classifiers of ad-hoc segmentation algorithms present in the literature will produce a reliable system, that can be used by physicians.