We propose a novel approach to the learning of conjoint acquisition and reconstruction of MRI scans. The acquisition is encoded in the form of general k-space trajectories, which constrained to obey the hardware requirements (peak currents and maximum slew rates of magnetic gradients). We demonstrate the effectiveness of the proposed solution in both image reconstruction and image segmentation, reporting substantial improvements in terms of acceleration factors and the quality of these end tasks. To the best of our knowledge, our proposed algorithm is the first to do data- and task-driven learning over the space of all physically feasible k-space trajectories.
This abstract and the presentation materials are available to members only; a login is required.