Vladimir Golkov1, 2, Tim Sprenger1, 3, Marion I. Menzel1, Ek T. Tan4, Kevin F. King5, Christopher J. Hardy4, Luca Marinelli4, Daniel Cremers2, Jonathan I. Sperl1
1GE Global Research, Garching n. Munich, Bavaria, Germany; 2Department of Computer Science, Technical University Munich, Garching n. Munich, Bavaria, Germany; 3IMETUM, Technical University Munich, Garching n. Munich, Bavaria, Germany; 4GE Global Research, Niskayuna, NY, United States; 5GE Healthcare, Waukesha, WI, United States
Sensitivity encoding (SENSE) reconstruction of diffusion weighted images (DWIs) in diffusion MRI is usually done independently for each DWI, without exploiting structural correlations between the DWIs. In this work, SENSE is incorporated into a joint reconstruction framework which models the prior knowledge of common smooth regions and edges in the DWIs. Image quality is improved in comparison to SENSE reconstruction, and even more so in combination with q-space compressed sensing.