Machine Learning Based Breast Cancer Detection

Abstract: In this study, a novel ultra-wideband (UWB) radar-based method for early, non-invasive breast cancer diagnosis is presented. The system uses a curved seventh derivative Gaussian UWB pulse that is filtered by a sharp transition bandpass FIR filter in both monostatic and bistatic radar configurations. The pulse shape enhances radiation power, spectrum efficiency, and FCC-compliant safety. Tumors are located and detected using Specific Absorption Rate (SAR) measurement and backscattered signal processing. Results from experiments and simulations confirm that the system can accurately identify cancers.