1411 - A Reinforcement Learning Approach for Optimization of Chemotherapy and its Application in Optimal Control

Abstract

ABSTRACT The diagnosis and treatment of the cancer has been the subject of discussion in different scientific fields in the recent years. In this paper, an optimal control strategy for the nonlinear systems is presented for application in chemotherapy of the cancer. The tumour growth model can be represented by a system of equations from population dynamics based on the competition between normal cells and tumour cells. The effect of the immune system on cancer is also included in the model. An optimal control method is applied to destroy the cancer cell with the minimum dose of chemotherapeutic agent. Reinforcement learning technique is proposed to construct an optimal control strategy for the nonlinear system and it was shown to be a suitable method to solve this problem. Simulation results show that cancer cells can be eradicated in a very short time with a small amount of drug using proposed optimal method of the therapy. They also show that the method is applicable to different models. It should be pointed out that the presented approach is a completely general procedure, and therefore, it can be applied to many problem of drug dosage optimization. Disclosure All authors have declared no conflicts of interest.