Network snakes for the segmentation of adjacent cells in confocal images

Abstract

Network snakes constitute one of the latest advances in the research of image segmentation techniques: they integrate topology into active contour models. This concept is applied to the problem of delineating biological cells in confocal images and is compared to results obtained with a current state-of-the-art segmentation method. Lacking a representative gold standard for this task, we adjust the well-established measure of segmentation quality developed by Pratt to our requirements and discuss the matching result of network snakes on one hand and conventional algorithms on the other. The work concludes with a discussion of some characteristic features and flaws of the approach.