Data-Driven Modeling and Optimization for Non-Invasive Microwave Hyperthermia

In this paper, the modeling for non-invasive microwave hyperthermia used in tumor treatment is introduced. By linearly simplifying the traditional Pennes model, we constructed two models to address the issue of unmeasurable internal temperature. The first model is the Output-Error(OE) model, which estimates the unmeasurable temperature inside the heated area based on measurable surface temperature. The second model is the Controlled Auto-Regressive(CAR) model, which describes the relationship between the microwave-driven signal and the estimated internal temperature. And they are identified respectively using the Particle Swarm Optimization (PSO) and the Mayfly Optimization Algorithm (MOA) intelligent optimization algorithms. Heating experiments were conducted on a biological platform, and the experimental results showed that the models under this modeling strategy have good accuracy and can provide good temperature estimations for microwave hyperthermia. Finally, experimental validation is conducted on a microwave hyperthermia platform. The proposed method exhibits excellent modeling accuracy, parameter convergence, and strong robustness, confirming its effectiveness for practical clinical medical applications.