Classification System for BilIN-1 and BilIN-2 based on Normal Nuclei-Array Model

Biliary intraepithelial neoplasm with intraductal papillary neoplasm of bile duct is known as a precursor lesion for intrahepatic cholangiocarcinoma. Therefore, to prevent the intrahepatic cholangiocarcinoma, it is important to diagnose the biliary intraepithelial neoplasm. The biliary intraepithelial is classified into BilIN-1, BilIN-2, and BilIN-3 by the atypia of cell and nuclei, loss of nuclear polarity, and disorder of the arrangement of nuclei. This paper classifies BilIN-1 and BilIN-2 using the proposed features based on the Normal Nuclei-Array Model with existing features for nuclei. For showing how the proposed features to improve the accuracy of classification, the experiment is performed by using the proposed features with the existing features of nuclei. The Classifier is used multilayer neural networks with back-propagation algorithm. The experiment results show that the accuracy of classification with the existing nuclei features and proposed features is improved about 13% than the classification with only the existing nuclei features.