Chronic diseases such as breast and ovarian cancer generally persist for a long time and cannot be easily prevented by vaccines or simply cured by medication. Such diseases require effective long-term monitoring, as well as periodic revisits for screening and treatment. Prophylactic surgeries involve removing organs from the patient's body would significantly reduce the risks of breast and ovarian cancers for mutation carriers. The problem is when should mutation carrier women receive prophylactic surgeries to not only decrease cancer incidence rate but also maintain a high level of quality-adjusted life years (QALYs). The optimal solution depends on sequential decision-making over long periods during the patient’s life involving the cancer-incidence probability, financial cost, and quality of life. The proposed research aims to develop a sequential decision-making framework for optimal cancer intervention strategies for chronic diseases through Monte Carlo Tree Search (MCTS). First, the progression dynamics of the disease under different medical interventions will be modeled as a continuous Markov Decision Process (CMDP). Second, the optimal intervention strategy will be solved via the MCTS algorithm, which increases computation efficiency by optimally balancing exploitation of current knowledge and exploration of uncertainty factors. The proposed framework will be further evaluated using real-world data.

Presenter(s)

Wuyang Qian