Investigation of Long Short -Term Memory Based Ultrawide Band Microwave Breast Tumor Size Prediction

Breast cancer is one of the most common malignancies among women in the world. Early detection and reliable monitoring are important factors in improving the survival rate and quality of life of cancer patients. The machine learning-based microwave tumor detection has received widespread attention because of its safety and cheapness. In current microwave breast cancer detection, most of the research is to determine the presence of tumors and few studies focus on monitoring the tumor development and estimate the size of breast cancer tumors. In this paper, by setting up an ultrawideband microwave detection system and collected sequence of microwave scattering signals of breast tissue, we estimate and monitor the tumor size by using the Long Short-Term Memory (LSTM) method. According to the Gompertz tumor growth model, we established 3D breast electromagnetic models to simulate the tumor growth process in Computer Simulation Technology (CST) electromagnetic simulation platform. The experimental results finally indicate that the LSTM algorithm can predict the size and growth trend of the tumor well and realize the monitoring of breast tumors.