Neural-network-based inverse-scattering technique for online microwave medical imaging

In this paper, the application of radial basis function neural networks (RBFNNs) in inverse-scattering problem related to microwave medical imaging is proposed. The objective of the networks is to estimate the geometric and/or electromagnetic properties of tissues by processing the scattered-field measurements obtained during the illumination of the body by electromagnetic waves. The training of the RBFNNs is based on the orthogonal least-squares algorithm. This approach results in straightforward construction of the network, where both the size and the values of the free parameters of the network are obtained. The proposed methodology is applied to the estimation of the position and the size of proliferated marrow inside the bone of a limp. This application is closely related to the detection and monitoring of leukemia. Different measurement configurations are examined. The performance of the constructed networks in cases of noisy field measurements is also investigated.