Convolutional Neural Network for Joint Detection and Material Classification of Breast Tumors

This paper presents a deep learning-based approach for comprehensively characterizing multiple breast tumors in a simple setup using microwaves. A convolutional neural network (CNN) detects and localizes the tumors, accurately determining their center coordinates and radii. Furthermore, the network is designed to estimate each identified object’s dielectric permittivity ($\epsilon_{\mathbf{r}}$), enabling classification into distinct material classes (0,1, and 2). The proposed methodology leverages the rich information content of microwave data to achieve robust and precise object detection and classification. Simulation results demonstrate the efficacy of the CNN in accurately extracting geometric parameters and material properties, showcasing its potential for applications in experimental breast cancer screening.