Target-driven acquisition and super-resolution reconstruction for magnetic resonance imaging

At the time of magnetic resonance imaging guided radiation therapy treatment delivery, the location of the target volume has been well established. During treatment of abdominal and thoracic tumors, real-time localization of the target using MRI is employed as a safety assurance measure. While high frame rate 3D imaging is desired for real-time localization, the inherently slow data collection speed of MRI limits the acquisition to planar 2D imaging intersecting the target. With this approach, the delivery can be paused if the target moves too far away from its ideal treatment position, but a volumetric history of the position of the target for each instance during delivery is lost. In this work, we develop a novel 2D multislice MRI encoding scheme in which every prescribed slice intersects the treatment target. With this approach, the rotating 2D slice always captures the position of the target for safety assurance, but a combined parallel imaging and super-resolution reconstruction can be used to produce 3D volumes in high detail near the target and lesser detail further away from the target. This novel encoding and reconstruction framework was demonstrated in phantom and in vivo experiments at 3T.