Automatic method for vessel detection in virtual slide images of placental villi

Abstract

The purpose of this work is to design an algorithm for automatic vessel recognition and counting on placental histological images in order to support the pathomorphological diagnostic. The studied images of placental villi come from spontaneous miscarriages and they are stained with Hematoxylin and Eosin. The proposed algorithm is based on colour component analysis, mathematical morphology operations, and decision tree classification. The major problems are variability of vessels and presence of collagen which can surrounds a villi. Based on the proposed method, automatic identification and counting of vessels is realized. The presented method can be applied as a support to traditional examinations.