An untrained deep learning method for reconstructing dynamic MR images from accelerated model-based data.
Kalina P SlavkovaJulie C DiCarloViraj WadhwaSidharth KumarChengyue WuJohn VirostkoThomas E YankeelovJonathan I TamirPublished in: Magnetic resonance in medicine (2022)
The use of an untrained neural network together with a physics-based regularization loss shows promise as a measure for determining the optimal stopping point in training without relying on fully-sampled ground truth data.