Brain Tumors Identification By Machine Learning-Based Quadratic Programming Born Iterative Method: A Preliminary Assessment

A Machine Learning approach is proposed in this work to solve inverse-scattering problems for brain cancer, by exploiting a well-assessed methodology from the same authors, yet applied for breast tumors. In particular, the Born iterative method with quadratic programming is first adopted to estimate the dielectric profiles from the measured scattered field. Then, a convolutional neural network with U-Net architecture is applied to successfully enhance the image reconstruction. The proposed approach is able to identify high-contrast brain tumors with greater accuracy than the original Born iterative method, as demonstrated by the reported numerical examples.