In recent years, computer-aided diagnosis based on machine learning, mainly based on Convolutional Neural Network (CNN) has been studied and developed rapidly. Those methods are not only helpful for classification, but also useful for feature extraction from given images, especially encoding image data into discrete representation helps us obtain new knowledge. Previous researches showed that CNN can by trainded for not only dection of cancers but also classification of gene expression subtypes. Although most of these studies are based on supervised learning that needs curated pathological knowledge, it is useful to extract characteristic features in the given images, using unsupervised machine learning in order to obtain new pathological findings. We applied cluster analysis using CNN which is trained based on adversarial training and maximization of mutual information and showed that it can classify those pathological images into discrete categories. Next, we applied our model for comparison of the two staining method in order to evaluate the degree of malignancy according to fibrosis and cell differentiation. The results showed that encoding of the histopathological image into discrete representations helps us to interpret tumor images.