Optimal dosing of cancer chemotherapy using model predictive control and moving horizon state/parameter estimation

Area of Science:

• Systems biology and control theory applications in oncology.
• Computational modeling for personalized cancer treatment strategies.

Background:

• Model predictive control (MPC) is a control strategy with potential for optimizing cancer therapy scheduling.
• Current MPC applications often assume full state observability, which is frequently not the case in biological systems.

Purpose of the Study:

• To demonstrate the utility of MPC for cancer treatment scheduling when system states are not fully measurable.
• To investigate the combined use of MPC with moving horizon estimation (MHE) to address state and parameter uncertainties.

Main Methods:

• Application of Model Predictive Control (MPC) integrated with Moving Horizon Estimation (MHE).
• Development of a state-space model to simulate tumor dynamics and drug effects.
• Simulation of tamoxifen dosing schedules for virtual tumor-bearing patients.

Main Results:

• The MPC-MHE scheme effectively schedules cancer therapy despite unmeasurable states and model parameter discrepancies.
• The study investigated the influence of estimation horizon length and the magnitude of parameter errors on treatment optimization.
• Simulations demonstrated the practical applicability of the combined MPC-MHE approach.

Conclusions:

• Model predictive control coupled with moving horizon estimation offers a robust framework for optimizing cancer therapy.
• This approach is effective even in the presence of significant model uncertainties and unobservable states.
• The findings support the use of advanced control strategies for personalized cancer treatment planning.