Adaptive therapy for non-small cell lung cancer via integrated Stackelberg game and deep reinforcement learning frameworks

Cancer is an evolutionary disease where the competitive evolution between sensitive cells and drug-resistant cells profoundly influences treatment outcomes. An in-depth investigation of cancer evolutionary dynamics holds significant importance for cancer therapy. This study constructs a polymorphic tumor evolution model based on Stackelberg game theory. We first conducted a systematic theoretical analysis to assess the model’s boundedness and the stability of its equilibrium points. Subsequently, leveraging clinical treatment data from non-small cell lung cancer patients, we implemented parameter calibration and validity verification to capture personalized characteristics of tumor evolution across different patients. Furthermore, we developed a personalized adaptive treatment strategy using deep reinforcement learning to formulate therapeutic regimens for patients with drug-resistant characteristics. Numerical simulation results demonstrate that this strategy precisely regulates the competitive balance between sensitive and resistant cells while ensuring patient quality of life, significantly suppressing resistance expansion and prolonging patient survival. This research provides novel insights and theoretical foundations for precision cancer treatment.