4-D PET-MR with Volumetric Navigators and Compressed Sensing

Abstract

Hybrid PET-MR scanners acquire multi-modal signals simultaneously, eliminating the requirement of software alignment between the MR and PET imaging data. However, the acquisition of high-resolution MR and PET images requires long scanning times, therefore movement of the subject during the acquisition deteriorates both the PET and the MR images. In this work we have developed an approach for tightly integrated PET-MR imaging, making use of volumetric MR navigators to inform, in real-time, both the MR acquisition and the PET reconstruction. The integrated imaging procedure that we describe exploits the simultaneity of MR and PET in hybrid PET-MR systems, producing inherently-aligned motion-free MR and PET images. We describe the system setup, the algorithm for motion-corrected reconstruction, an adaptive sinogram binning algorithm and software design decisions aimed at integrating tightly the MR and PET subsystems. Application of the integrated motion-corrected acquisition procedure to a phantom study and to a volunteer subject demonstrates the validity of the approach for a variety of motion patterns.