Computational MRI

Novel sampling and reconstruction methods: from accelerated k-space acquisition to diagnostic-quality image

We develop advanced sampling and reconstruction methods for ultrafast dynamic MRI, overcoming traditional trade-offs between spatial resolution, temporal resolution, and scan time. Our methods, such as ELITE, combine locally low-rank subspace modeling, compressed sensing, and deep learning to suppress undersampling artifacts while preserving temporal fidelity. These techniques enable high temporal-resolution imaging that improves lesion characterization, reduce noise, and support reliable quantitative kinetic analysis in applications such as dynamic contrast-enhanced (DCE) breast and abdominal MRI.