Motivation: To meet the clinical demand for fast and motion-robust brain MRI. Goal(s): To integrate retrospective motion correction with a data-driven deep learning reconstruction method to achieve high-quality, motion-robust 2D TSE imaging. Approach: Motion trajectory information was derived from scout and guidance line-based motion correction. A data-driven deep learning network was developed, interleaving multiple conjugate gradient SENSE (+motion) optimizations with network regularization, and was trained and evaluated on TSE data. Results: The method demonstrated improved signal-to-noise ratio (SNR) and reduced motion artifacts in vivo, utilizing 4-fold accelerated scans with induced step motion. Impact: We integrate retrospective motion correction into a data-driven deep learning network to facilitate fast and motion-robust 2D TSE imaging in the brain.
Lugo et al. (2025) studied this question.