Recent development of simultaneous whole-body PET-MR scanners has the potential of performing non-invasive diagnosis of coronary artery disease, including assessment of myocardial perfusion, coronary angiography and ventricular function in one comprehensive exam¹. However, image quality degradation due to physiological motion of both PET and MR images remains a major challenge. For cardiac MR imaging, image-based navigators have been proposed in order to address the problem of respiratory motion for 3D whole heart coronary MR angiography (MRA)^2-3. These approaches reduce the MR acquisition time, since all acquired data is used for reconstruction (100% scan efficiency) and correct for the complex motion of the heart during free-breathing acquisition. One of these approaches spatially encodes the start-up echoes of a balanced steady state acquisition sequence to reconstruct an image navigator². Here we propose to extend this approach to a gradient echo sequence, so that start-up echoes can be used to acquire image-based navigators for inline 2D translational motion corrected coronary MRA imaging in a 3T PET-MR system.

An ECG-triggered 3D T1-weighted gradient echo sequence using a fully sampled golden-step Cartesian spiral profile order⁴ was implemented. This trajectory samples the phase encoding plane following approximate spiral interleaves on the Cartesian grid. A user-defined number of start-up echoes were used to acquire a coronal-oriented low-resolution 2D Cartesian navigator at each heartbeat. Additionally, an adiabatic T2 preparation pulse was implemented to improve the contrast in the images, and a fat saturation prepulse was applied before 3D coronary MRA acquisition (Fig. 1). Translational motion in foot-head (FH) and left-right (RL) directions was estimated from the 2D navigator in a beat-to-beat fashion using normalised cross correlation of a template covering the apex of the heart (Fig. 2). Motion estimates were used to correct data acquired at each heartbeat prior to image reconstruction by multiplying the k-space data with corresponding linear phase factors. Inline motion correction was implemented in the scanner, so that a high-resolution 3D motion corrected coronary MR angiography was obtained. Acquired data was also reconstructed inline without motion correction for comparison purposes.

Three healthy subjects were scanned during free-breathing on a 3T PET-MR scanner (Biograph mMR, Siemens Healthcare, Erlangen, Germany) using a prototype implementation of the proposed gradient echo sequence (field of view = 300x300x80mm³, resolution = 1x1x2mm³, TR/TE = 3.75/1.72ms, flip angle = 15°). A subject-specific acquisition window (82–90ms) was obtained by varying the segments acquired per each spiral interleave (22-24 profiles per cardiac cycle), and a trigger delay was set targeting the mid-diastolic rest period. For the 2D image navigator acquisition, 14 start-up echoes (same FOV, flip angle = 3º) were used.

3D coronary MRA images were reformatted to simultaneously visualize the left anterior descending (LAD) and right coronary artery (RCA) using custom made software⁵. A significant improvement in recovering the distal segment of the RCA (red arrow) and the proximal segment of the LAD (blue arrow) can be observed in Fig. 3 (top) for the first subject. However, lack of contrast prevented to observe the distal LAD. A similar behaviour can be observed in Fig. 3 (bottom) for a second subject. Additionally, blurring of the RCA (green arrow) is reduced when applying the proposed motion correction.We have presented an MR acquisition scheme that allows for inline translational motion correction for coronary MR angiography in a 3T PET-MR scanner. By spatially encoding the start-up echoes of a gradient echo sequence, motion was estimated in a beat-to-beat fashion, achieving 100% scan efficiency. A golden-step Cartesian spiral profile order acquisition was used for acquiring high-resolution 3D MR data. Motion correction improves image quality and contrast compared with uncorrected images. Further work includes incorporating non-rigid bin-to-bin motion correction to further improve image quality, and use of the MR-measured motion estimates to correct simultaneously acquired cardiac perfusion PET data and corresponding attenuation correction maps.