Breast Cancer Detection Using Microwave Signal and Classification Using SVM

In this chapter, breast cancer detection using microwave signals and classification using SVM (Support Vector Machine) has been studied. Two microstrip patch antennas calibrated at 2.45 GHz are designed and placed on opposite sides of the 3D breast phantom model in Ansys HFSS. The breast phantom model has been designed for three classes: normal breast, benign tumor, and malignant. The scattering parameters, \(S_{11}\) and \(S_{21}\) , has been collected for all three classes. Collected scattering parameters were converted to the time domain from the frequency domain. After the conversion, a four-level DWT (Discrete Wavelet Transform) feature extraction with db6 as mother wavelet is applied, and time-frequency features were extracted and concatenated. The concatenated time-frequency features of both \(S_{11}\) and \(S_{21}\) parameters were used to train the SVM model to give an accuracy of 88.33% and 90% for unaugmented and augmented data, respectively.