Deep Learning Strategies for Quantitative Biomedical Microwave Imaging

This paper proposes a numerical performance assessment of the recovery capabilities in the framework of microwave imaging via deep learning approaches. More in detail, aim of the analysis is the comparison among different convolutional neural network architectures in order to understand the impact of each parameter on the recovery performance for quantitative imaging. To support the analysis, some quality metrics were evaluated and a comparison with a conventional nonlinear approach is considered. The results seem promising, both in terms of computational time and recovery accuracy, especially in very noisy scenarios with a limited amount of data.