Survival Prediction of Bladder Cancer Based on Weakly Supervised Learning

Currently, bladder cancer is one of the top 10 most prevalent cancers among Chinese men. In this study, we propose a weakly supervised learning model, i.e., WSI-SPB model, for survival prediction of bladder cancer. The proposed WSI-SPB model integrates features exploited from clinical data and the whole slide images of bladder cancer patients, which develops an innovative way for establishing imaging indicators and achieves improved results. It accomplishes automatic survival prediction analysis without the necessity of image annotation by medical professionals. Notably, the optimal model in this experiment achieved an AUC of 0.80, with a recall of 90.91% among the prediction indices. The model developed in this paper holds the potential to aid doctors in making intelligent predictions in clinical practice, facilitating the formulation of patient follow-up, diagnosis, and treatment plans, ultimately enhancing work efficiency.