Pre-treatment mapping strategies to enhance real-time temperature reconstruction in microwave cancer hyperthermia

The present thesis focuses on the development of error reduction methods in simulation-guided hyperthermia (HT) treatments, in particular for head and neck (H&N) tumors. HT¬¬¬¬ therapy has gained increasing attention in the medical field due to its role in enhancing conventional cancer treatments such as radiotherapy and chemotherapy. By elevating the temperature of tumor tissues to 40-44 °C using non-ionizing microwave radiation, it increases the sensitivity of cancer cells to radiation and improves drug delivery without introducing additional toxicity. The benefits of HT have been well-demonstrated across various tumor sites, including cervix, breast, head and neck, skin, bladder, and esophagus. For sub-superficial and deep-seated tumors, phased-array antenna systems are used to optimize the Specific Absorption Rate (SAR) within the tumor target, while minimizing the risk of hotspots in the surrounding healthy tissues. Following clinically prescribed hyperthermia treatment planning (HTP) guidelines, patient-specific numerical simulations and temperature monitoring systems play a crucial role in controlling active electronic systems that adjust antenna feedings during treatment sessions while simultaneously monitoring the achieved temperature in different regions. The research particularly focuses on exploring pre-processing strategies to improve the real-time reconstruction of the patient’s temperature distribution from limited invasive measurements. Chapter 1 provides a comprehensive overview of hyperthermia, covering its historical background, biological effects, and the mathematical modeling of tissue heating, highlighting the limits of the Pennes bioheat equation and the uncertainties associated with tissue properties, which can significantly affect the simulation outcome. Chapter 2 delves into the challenges associated with current temperature monitoring techniques, such as the invasiveness of thermometry catheters. The chapter introduces the high-resolution virtual model used in this study and the concept of parameter space. The procedures for electromagnetic and thermal simulations are detailed, as well as the method used to predict the patient temperature distribution from limited measurement data, which relies on inversion algorithms. Chapter 3 investigates the integration of S-parameters (active reflection coefficients) into the reconstruction process to enhance the accuracy of the predictions. The impact of this approach in different scenarios is evaluated using three metrics: the minimum error threshold in 95% of the volume and in the 95% of the tested cases, the median of the pointwise absolute difference between the reconstructed and actual temperature map, and the absolute difference between the reconstructed T50 parameter in the tumor volume and the actual T50 in the tumor volume. Chapter 4 focuses on optimizing the exploration of the parameter space by initially employing Sobol sequences to ensure uniform coverage, followed by targeted refinement in specific regions. Three types of refinement were investigated: two aimed at reducing redundancy in the basis matrix and one targeting underrepresented regions associated with higher errors. This approach seeks to identify the most effective sampling strategies to minimize error magnitude and occurrence in both tumor and healthy tissues. The results of this analysis were reported in terms of error percentiles and the considered metrics.