Automatic segmentation of corneal endothelial cells using active contours

We present a new method for segmenting the corneal endothelial cells from microscopic images. It uses multiple active contours initialized by adaptive thresholding and limited with their growing to not overlap. Thanks to the inherent characteristics of the active contour both outcomes can be achieved: cell quantity and delimitation. The tool implementing this approach is built within the MESA framework an environment for developing and evaluating segmentation techniques. The accuracy is estimated on the base of real microscopic cell images segmented manually.