K-nearest unrepeatable cell graph model of histopathological tissue image

One of the most important components in the histopathological tissue images is the cell nuclei. Features such as the number, morphological properties and location of the cell nuclei offer useful information for histopathological analysis. Cell-graph models are constructed using location information of cell nuclei and important distinctive information can be obtained from the features of the models. The models are generally formed according to the distance between the cell nuclei. However, the distance between the cell nuclei is affected by various factors during obtaining tissue image and shows variety. In this study, using one-way neighborhood relationship of the nuclei with each other is proposed for the construction of the cell-graph models of histopathological images. The proposed approach has been tested on 20 healthy and 20 necrotic liver tissue images. The results show that graph models constructed by the neighborhood relationship, have more distinctive characteristics than distance-based graph models.