A Real-time Breast Hyperthermia Monitoring Scheme Based on Processing of Microwave Scattering Parameters with Deep Learning

Abstract

This study proposes a new method based on deep learning to determine whether the temperature values ​​are at an appropriate level during the use of microwave hyperthermia method in the treatment of breast cancer. To implement our method, we utilize the temperature dependent dielectric properties of biological tissues to generate the heating scenarios that simulates the thermal behavior of biological tissue during the breast cancer hyperthermia treatment. Using the temperature-dependent dielectric properties we designated corresponding temperature thresholds, next, we labeled the malignant tumor region and the healthy tissue region in accordance with the pre-determined thresholds. In addition, scattering problems are solved based on treatment (hot or heated) and pre-treatment (cool) scenarios. Using the difference between hot and cool states, we train, test, and validate the CNN. Our main purpose in the project is to determine whether the tissue is heated in the desired temperature region using only the single frequency differential scattered electric field data.

Microwave Antenna Excitation Optimization with Deep Learning For Breast Cancer Hyperthermia

“…Deep learning has been employed as a tool for solving the medical diagnostics and therapeutics problems in the past decade such as registration of multi modality diagnostic images [20], tissue type classification [21], optimization of elements in ultrasound thermometry [22], MH monitoring [23], and many more, and it continues to extend its validity. One of the indispensable applications of deep learning is the antenna array selection and excitation optimization problem in beamforming which is a high-cost problem [24]- [26].…”