Closing the Gap of Simulation to Reality in Electromagnetic Imaging of Brain Strokes via Deep Neural Networks

Bringing deep learning techniques to electromagnetic imaging is of interest considering its great success in various fields. Deep neural nets however are known for being data hungry machines, and in many practical cases, such as electromagnetic medical imaging, there is not enough to feed them. Scarcity of data necessitates reliance on simulations to generate a sufficiently large dataset for deep learning to perform any complicated task. Simulations however, can not perfectly represent real environments and therefore, any neural net trained on simulation data will invariably fail when evaluated on real data. This work customizes a deep domain adaptation technique for matching distributions of complex-valued electromagnetic data. We demonstrate the advantage of using complex-valued models over regular ones. An operational neural network trained on simulation data and adapted to practical data to perform brain injury localization is presented.