Feasibility Study of Breast Cancer Detection Using Microwave Imaging with the Machine Learning Approach

Microwave imaging is widely spreading in the field of biomedical diagnostics due to its efficient feature of contrast detection. Over a decade, it is prominent in the field of breast cancer detection. Microwave imaging comes up as the non-invasive and non-ionizing approach with least discomfort to the patient in comparison to other conventional approaches. The tumors typically exhibit a higher dielectric value compared to the surrounding breast tissues. Consequently, microwave imaging can facilitate the detection and localization of malignant tumors. The dielectric properties of breast tissue exhibit dispersive characteristics, making it important to estimate these properties for evaluating the efficacy of microwave imaging systems. Additionally, the absence of distinct separation between individual layers of breast tissue necessitates the determination of the effective dielectric constant. Thus, a dielectric mixing model is proposed to estimate the dielectric constant of the breast tissues efficiently. This study mainly focuses on the determination of the feasibility of microwave imaging in the 1–10 GHz frequency range. The critical analysis of the electrical properties of the breast tissues with and without tumor is done in this proposed work. Furthermore, synthetic data generation is done and analyzed using the Support Vector Machine (SVM) algorithm for the classification of the tumor.