Simulating chemotherapy of tumor models via adopting an artificial intelligence approach

Abstract Cancer is a major global health challenge, urgently requiring innovative treatment strategies. Traditional chemotherapy regimens often lack adaptability to the dynamic tumor microenvironment. In recent years, artificial intelligence (AI) technologies, particularly reinforcement learning, have provided novel tools for optimizing personalized therapies. This study presents a novel AI-based framework utilizing the Proximal Policy Optimization (PPO) algorithm to simulate and optimize chemotherapy protocols within tumor models. The framework employs a system of dynamic differential equations to model the complex interactions among tumor cells, normal cells, and immune cells, with drug concentration serving as the primary control variable. By interacting with the tumor model environment, the PPO algorithm generates optimized drug administration strategies that simultaneously minimize tumor cell proliferation and mitigate drug-induced side effects. Experimental results demonstrate the efficacy of this approach in controlling tumor growth across both noise-free and noisy environmental conditions. Moreover, introducing noise into the training process notably improves the model’s resilience and its ability to adapt to varying conditions.