Fast and Accurate CNN-Based Machine Learning Approach for Microwave Medical Imaging in Cancer Detection

The application and comparison of U-Net convolutional neural network architectures is proposed in this work to guarantee a fast accurate convergence inverse scattering problems solved by Born iterative method, even the presence strong scatterers. Starting from preliminary configuration authors some recent papers, two variants are introduced discussed significantly reduce computational cost, while guaranteeing with very high accuracy dielectric profiles reconstruction when considering scattereres, such as tumors, thus working regularization process mitigate induced non-linearity. As further enhancement, novel approach which integrates Resnet models realize segmentation process, leading effective feature extraction identification anomalies within healthy tissue. Numerical assessments on variety breast including abnormal lesions successfully validate machine learning approach, through adoption properly defined evaluation metrics.