A Stochastic Model Predictive Control Approach to Deal with Cancerous Tumor Growth

Tumor growth models can help predict the response of tumors to different treatments. This work presents a generic mathematical model that combines tumor growth, the pharmacokinetics of the drugs administered, and the evolution of the possible side effects of the treatment. Tumors are complex systems, and they can exhibit different growth patterns under the same initial conditions. To deal with these uncertainties, a stochastic model predictive strategy has been developed to reduce the tumor size while minimizing side effects. Thus, probabilistic constraints have been incorporated into the optimization problem, giving rise to a chance-constrained model predictive approach. A one-year treatment simulation assessment is performed to compare the stochastic model predictive controller with the standard implementation. The results demonstrate that the proposed approach improves the performance of the controller, satisfying the marked objectives. Overall, while tumor growth modeling can provide valuable insights into the behavior of tumors, it is important to incorporate sources of uncertainty to ensure that the models accurately capture the behavior of all tumors.