Unsupervised subtyping of cholangiocarcinoma using a deep clustering convolutional autoencoder

Unlike common cancers, such as those of the prostate and breast, tumor grading in rare cancers is difficult and largely undefined because of small sample sizes, the sheer volume of time and experience needed to undertake such a task, and the inherent difficulty of extracting human-observed patterns. One of the most challenging examples is intrahepatic cholangiocarcinoma (ICC), a primary liver cancer arising from the biliary system, for which there is well-recognized tumor heterogeneity and no grading paradigm or prognostic biomarkers. In this paper, we propose a new unsupervised deep convolutional autoencoder-based clustering model that groups together cellular and structural morphologies of tumor in 246 digitized whole slides, based on visual similarity. Clusters based on this visual dictionary of histologic patterns are interpreted as new ICC subtypes and evaluated by training Cox-proportional hazard survival models, resulting in statistically significant patient stratification.

• View at Publisher
• View at Scopus