End-to-End Sequential Sampling and Reconstruction for MRI
Accelerated MRI shortens acquisition time by subsampling in the measurement $κ$-space. Recovering a high-fidelity anatomical image from subsampled measurements requires close cooperation between two components: (1) a sampler that chooses the subsampling pattern and (2) a reconstructor that recovers images from incomplete measurements. In this paper, we leverage the sequential nature of MRI measurements, and propose a fully differentiable framework that jointly learns a sequential sampling policy simultaneously with a reconstruction strategy. This co-designed framework is able to adapt during acquisition in order to capture the most informative measurements for a particular target. Experimental results on the fastMRI knee dataset demonstrate that the proposed approach successfully utilizes intermediate information during the sampling process to boost reconstruction performance. In particular, our proposed method can outperform the current state-of-the-art learned $κ$-space sampling baseline on over 96% of test samples. We also investigate the individual and collective benefits of the sequential sampling and co-design strategies.