A non-invasive microwave imaging system with step-by-step tumor localization for early breast cancer diagnosis

Breast cancer is one of the most common cancers in women and has a relatively low mortality rate. Early detection is essential to improve survival rates. Currently, conventional techniques for breast cancer detection expose patients to harmful radiation and require breast compression, which is uncomfortable. Therefore, microwaves have emerged as a promising technique for early detection because they are safe and non-invasive, enabling early detection of malignancy. In this work, an ultra-wideband antenna is designed using high-frequency simulation software (HFSS). The backscattered signals from the antenna, when placed around the breast, are processed with a confocal imaging algorithm to locate the tumor. This study presents the imaging algorithm in detail, step by step. Additionally, a Support Vector Regression algorithm (SVR) is applied to the dataset obtained from the antenna to estimate detection accuracy. The SVR algorithm achieves a 95.94% accuracy rate using statistical feature extraction with a radial basis function kernel for tumor detection.