Abstract
644
Objectives Our aim is to develop a fast 3D MR acquisition for real-time tracking of 3D respiratory motion (RM) during a simultaneous PET-MR scan that will enable RM-corrected attenuation correction and PET reconstruction. Existing MR-based techniques derive 3D non-rigid motion from composite 3D images reconstituted by sorting multi-slice 2D MR images acquired over many respiratory cycles [1,2,3], requiring separate respiratory gating and are subject to inter-slice misalignment. Our real-time technique can measure RM at all times and only use MR data acquired for imaging.
Methods An in-house fast 3D radial MR pulse sequence was developed. Spoiled gradient-echo scans were performed on a free-breathing healthy volunteer on a 3T Siemens Magnetom Trio scanner (similar to MR unit of the Siemens Biograph mMR). Over a 22-sec scan, 12,000 radial readouts (128 samples each over 400mm field of view) were arranged using multidimensional golden means [4] to allow flexible frame rate in retrospective reconstruction.
Results Raw MR data was reconstructed to various spatiotemporal resolutions. A sliding window scheme of 1,000 readouts per window and 50% overlap resulting in 1 image/sec and 3.1mm isotropic voxels was found to effectively visualize the lung-liver interface and 3D RM in all planes and phases. Spatiotemporal resolution and signal-to-noise ratio (S/N) are expected to improve as we apply advanced reconstruction techniques that leverage the spatiotemporal sparsity of RM.
Conclusions We developed a method for real-time MR tracking of 3D non-rigid RM for diagnostic MR imaging in simultaneous PET-MR. This approach is suitable for deriving 3D motion field for PET reconstruction, especially when PET S/N ratio is too low for RM correction using PET data itself as in our previous works.