Dynamic Bayesian networks for prediction of health status and treatment effect in patients with chronic lymphocytic leukemia.

Chronic lymphocytic leukemia (CLL) is the most common blood cancer in adults. The course of CLL and patients' response to treatment are varied. This variability makes it difficult to select the most appropriate treatment regimen and predict the progression of the disease. This work was aimed at developing and validating dynamic Bayesian networks (DBNs) to predict changes of the health status of patients with CLL and progression of the disease over time. Two DBNs were developed and implemented i.e. Health Status Network (HSN) and Treatment Effect Network (TEN). Based on the literature data and expert knowledge we identified relationships linking the most important factors influencing the health status and treatment effects in patients with CLL. The developed networks, and in particular TEN, were able to predict probability of survival in patients with CLL, which was in line with the survival data collected in large medical registries. The networks can be used to personalize the predictions, taking into account a priori knowledge concerning a particular patient with CLL. The proposed approach can serve as a basis for the development of artificial intelligence systems that facilitate the choice of treatment that maximizes the chances of survival in patients with CLL.