A Systematic Approach for Glandular Structure Segmentation from Colon Histopathology Images

In this paper a method is proposed for segmenting glandular structures from Haematoxilyn and Eosin stained colon histopathology images. Gland includes three main structures: lumen, cytoplasm and cell nuclei. As the first step of the algorithm, k-means clustering algorithm is applied to cluster images to 3 clusters (lumen, cytoplasm and cell nuclei). Then, all lumen in each histopathology images are extracted. A set of textural and structural features are extracted from each lumen. Cytoplasm that are adjacent to each lumens are detected. Then a set of textural and structural features are extracted from each cytoplasm. At the next step, cell nuclei rings that are adjacent to each cytoplasm are detected. Then a set of textural and structural features are extracted from each cell nuclei rings. Three feature that extracted from lumen, cytoplasm and cell nuclei rings are combined as a feature vector. The extracted features are then fed into an SVM classifier with a RBF kernel. The experimental results show good capability of the proposed method for gland segmenting from colon histopathology images.