Susceptibility-based Positive Contrast Imaging of MR Compatible Metallic Devices Using Model-based Deep Learning

Susceptibility-based positive contrast MR imaging exhibits excellent efficacy for visualizing the MR compatible metallic devices, by taking advantage of their high magnetic susceptibility. In this work, a model-based deep learning architecture with U-net is developed to realize the 3D susceptibility-based positive contrast MR imaging on real phantom experiments. We train the network on synthetic data to generate positive contrast images from magnetic field maps for localizing the seeds from their surroundings and demonstrate the potential of the deep learning implementation.