Respiratory motion-compensated ROCK-MUSIC with RF receiver cardiac focusing in pediatric congenital heart disease (RF_FR_403)

Rotating Cartesian k-space sampling with ferumoxytol enhancement (ROCK-MUSIC) is a self-gated (SG) MRI pulse sequence, developed for fast whole-heart 4D imaging in pediatric congenital heart disease (CHD).¹ It currently requires general anesthesia and controlled mechanical ventilation. Recent work has shown that low-rank tensor (LRT) reconstruction² of ROCK-MUSIC data enables higher temporal resolution and respiratory motion-resolved imaging.³ Further improvements in handling respiratory motion are desired for free-breathing (FB) ROCK-MUSIC without anesthesia.

Here, we propose rigid respiratory motion-compensated (MoCo) LRT image reconstruction with retrospective cardiac phased array RF focusing using region-optimized virtual coils (ROVir)⁴. Rigid MoCo alone does not remove respiratory motion from the LRT temporal model. Thus, we hypothesized that applying ROVir prior to temporal basis estimation would produce a more efficient, cardiac-focused temporal basis, improving the representation of cardiac motion and reducing respiratory motion artifacts.

Methods:

Ten CHD patients (3 days–6.7 yrs) were scanned on a 3T Siemens Trio using ROCK-MUSIC, a 3D gradient-recalled-echo sequence with TE/TR=1.2ms/2.9ms, isotropic spatial resolution=0.8-1.1mm³, flip angle=19°, during the steady-state intravascular distribution of ferumoxytol, 4mg/kg.¹ LRT reconstruction used SG lines for temporal basis estimation and for cardiac/respiratory binning, reconstructing 5D images with 20 cardiac bins × 6 respiratory bins³. Rigid respiratory MoCo calculated bin-to-bin image translations minimizing the nuclear norm over a cardiac ROI then directly compensated kt-space data, including the SG lines. ROVir was implemented prior to LRT image reconstruction (Fig. 1) to learn a cardiac-focused temporal basis.

We evaluated MoCo+ROVir’s ability to focus on cardiac rather than respiratory motion by comparing: 1) preservation of static, cardiac, and respiratory energy 2) cardiac/respiratory energy ratio in the SG lines used for temporal basis estimation. We further evaluated flickering and residual motion artifacts in the reconstructed images by comparing 3) dynamic/static energy ratio over the arm. We compared the original LRT reconstruction (No-MoCo), MoCo and MoCo+ROVir.

Results:

MoCo compensated respiratory motion over the heart but introduced spurious reverse motion in static tissues, whereas MoCo+ROVir reduced reverse motion and respiratory artifacts (Fig. 2). ROVir preserved more cardiac than static and respiratory energy (p< 0.001), increased the cardiac/respiratory energy ratio (p< 0.001) prior to temporal basis estimation, and reduced flickering artifacts (p< 0.001) (Fig. 3).

Conclusion:

MoCo and ROVir was evaluated in LRT reconstruction of ROCK-MUSIC data for pediatric CHD, successfully isolating cardiac motion and reducing respiratory artifacts. These techniques show promise for FB ROCK-MUSIC in pediatric CHD, potentially eliminating the need for general anesthesia. Further validation including quantitative cardiac function evaluation in a larger cohort is warranted.