Serial MRI, where patients undergo repeated imaging over time, is standard for tracking disease and for procedures like MR-guided radiotherapy. But each scan takes time, and that adds up: longer exams mean higher costs and reduced access for patients. The new arXiv preprint, LGGS, tackles this bottleneck by focusing on reconstruction rather than hardware.
LGGS stands for Longitudinal MRI Reconstruction via Guided Group Sparsity. The idea is to reconstruct high-quality images from undersampled data, using the fact that in a longitudinal series, consecutive scans share a great deal of structure. By guiding the reconstruction with group sparsity—a mathematical way to enforce shared patterns across images—the method aims to produce diagnostically useful images from faster acquisitions.
If the approach holds up, it could make serial MRI more practical in busy clinical settings, particularly for surveillance and adaptive radiotherapy, where speed and consistency matter. The paper is a single source, so there are no conflicting findings to compare; the abstract itself does not report specific performance numbers.