In conventional prospective respiratory navigator (NAV) acquisitions, 40-60% of acquired data are discarded due to respiratory motion, resulting in low efficiency and long scan times. Compressed-sensing Motion Compensation (CosMo) aims to reduce scan time by estimating missing k-space lines. This study investigates whether incorporating the magnitude of motion-corrupted (rejected) k-space lines as a constraint in image reconstruction can improve CosMo performance. Results from 10 healthy subjects demonstrate that this additional constraint significantly improves the signal-to-noise ratio (SNR) and image quality of right coronary artery (RCA) imaging without increasing scan time.
Introduction
Imaging the heart requires managing both cardiac and respiratory motion. Standard navigator gating (NAV) only records data when the diaphragm is within a narrow (typically 5mm) window. This leads to substantial data waste and prolonged scan durations. While methods like CosMo use compressed sensing to fill in missing data, they often overlook the information contained in the "rejected" motion-corrupted lines. Since respiratory motion primarily affects the phase of k-space data while leaving the magnitude relatively stable, these rejected lines can serve as a valuable anchor for reconstruction.
Methods
To evaluate magnitude stability, free-breathing, ECG-triggered RCA images were acquired from 10 healthy volunteers. Magnitude variability was quantified via cross-correlation between accepted and rejected k-space lines.
The study implemented CosMo retrospectively using the Likelihood Operator for Sparse Transformation (LOST) algorithm. A new magnitude constraint was introduced, where the magnitude of reconstructed lines was required to match the magnitude of the rejected lines already stored in memory. Two variations were tested: applying the constraint within every iteration and applying it only in the final iteration. These were compared against standard 5mm NAV gating and the original CosMo reconstruction.
Results and Discussion
Cross-correlation analysis confirmed that magnitude information remains highly consistent across respiratory phases, especially at the center of k-space (correlation ~1.0).
Figure 2a: Right coronary artery image acquired using standard fully-sampled, 5-mm gated data (baseline).
Visual and quantitative assessments showed that the magnitude constraint significantly enhanced image clarity. The relative SNR in the left ventricle blood pool improved from 30.71 ± 6.5 with standard gating to 56.8 ± 25.9 when using the magnitude constraint in the final iteration.
Figure 2d: Right coronary artery image using CosMo with the additional magnitude constraint applied in the final iteration, showing the highest SNR and improved vessel definition.
Statistically significant differences (p < 0.05) were observed between the proposed method and both the standard gating and original CosMo reconstructions. The visual results show sharper vessel edges and reduced background noise, particularly in the version where the constraint is applied in the final iteration.
Conclusion
Incorporating the magnitude of rejected k-space lines—data that is already acquired but typically discarded—provides a "zero-cost" improvement to cardiac MRI. This additional constraint in the CosMo framework results in superior image quality and higher SNR, offering a promising path toward more efficient and reliable coronary artery imaging.
Summary updated
Abstract
Background. In conventional prospective respiratory navigator (NAV) acquisitions, 40-60% of the acquired data are discarded resulting in low efficiency and long scan times [1,2].Compressed-sensing Motion Compensation (CosMo) has a shorter fixed scan time by acquiring the full inner k-space and estimating the NAV-rejected outer k-space lines [3]. Respiratory motion will mainly manifest itself as phase variation in the acquired k-space data. We sought to determine if the addition of the magnitude of the rejected k-space lines as a constraint in image reconstruction will improve the performance of CosMo. Methods. To investigate the variability of the magnitude of kspace lines at different respiratory phases, free-breathing, ECG-triggered, targeted right coronary images with multiple averages were acquired from 10 healthy adult subjects. Magnitude variability was investigated quantitatively by calculating the cross-correlation between accepted and rejected k-space lines. CosMo was implemented retrospectively on one acquisition from each subject. The inner k-space (31 ky by 7 kz lines) was filled with lines acquired within the 5mm gating window from all acquisitions. The outer kspace was then filled only with lines from the first average acquired within the 5 mm gating window, resulting in an undersampled k-space with a fully sampled center. For reliable image reconstruction with CosMo, 10-20% of the inner k-space must be fully-sampled. The missing outer k-space lines were then estimated using LOST with an additional magnitude constraint within each estimation iteration or in the final iteration for each coil [4]. The results were compared with prospective NAVgating with a gating window of 5 mm and CosMo reconstruction without the magnitude constraint. Results. Figure 1 shows the cross-correlation between the accepted and worst rejected k-space lines for each position. The correlation is close to 1 at the center of kspace where the majority of image information is contained, indicating low variability in magnitude information at different respiratory phases. Figure 2 shows right coronary images acquired using a) fully-sampled, 5-mm\ngated data, b) the original CosMo, and CosMo with the additional magnitude constraint c) inside each iteration and d) in the final iteration. The relative signal-to-noise in the left ventricle blood pool is: \\(30.71 \\pm 6.5; 40.32 \\pm 14.2; 53.9 \\pm 26.8; 56.8 \\pm 25.9\\) for each reconstruction, respectively. Significant differences (p<0.05) are present for all measurements except between the original CosMo and the CosMo image with the magnitude constraint in each iteration (p=0.09). Conclusions. The addition of the magnitude of rejected lines, readily available in all navigator-gated scans, as a constraint in CosMo results in improved image quality as measured by relative SNR. Funding. NIH R01EB008743-01A2.
BibTeX
@article{shaw2012improved,
author = {Jaime L. Shaw and Mehdi H. Moghari and Mehmet Akçakaya and Raymond H. Chan and Warren J. Manning and Reza Nezafat},
title = {{Improved navigator-gated motion compensation in cardiac MR using additional constraint of magnitude of motion-corrupted data}},
journal = {Journal of Cardiovascular Magnetic Resonance},
year = {2012},
doi = {10.1186/1532-429x-14-s1-p245},
}