Deep learning for histopathology of the lower urinary tract

Histopathology is the cornerstone in the diagnosis and treatment-decision making for many cancers. The examination consists of scoring two aspects of the tumor, the tumor aggressiveness (grade) and the tumor location and spread (stage). A significant drawback of this examination is the high inter-observer variation in both grade and stage of the tumor, potentially leading to suboptimal treatment.
In this thesis, we try to improve the histopathological grade examination for prostate and bladder cancer by using deep learning. In this case, automated decision making tools are trained on histopathology slides labeled by pathologists. Nonetheless, these techniques are still subjected to the high inter-observer variation, and to that end we also try to automatically predict the long-term outcome of bladder cancer patients.