Development of a metric and the methods for quantitative estimation of the segmentation of biomedical images
Oleh Berezsky, Mykhailo Zarichnyi, Oleh Pitsun · Eastern-European Journal of Enterprise Technologies · 2017
We analyzed modern digital microscopy. In order to categorize digital microscopy, the following criteria are introduced: level of automation, software level, the level of application of network technologies. To quantitatively estimate the quality of image segmentation, we devised the technique based on a metric approach using the Fréchet and Hausdorff metrics. Modern algorithms for calculating the Hausdorff and Fréchet distances were analyzed.We have introduced the Fréchet distance between trees. It was proven that the Fréchet distance between trees is a metric. We devised a method for estimating a distance between trees of the non-convex regions, based on finding skeletons of regions and determining the distance between them. The algorithm for finding the Hausdorff distance between the non-convex regions is described. We constructed the algorithm for finding a distance between the non-convex regions based on the Fréchet metric between trees.The developed algorithms are included into a hybrid intelligent system for automated microscopy, which is designed to process histological and cytological images.The algorithms were tested using the results of segmentation of histologic and cytologic images from a database as an example