Abstract
To evaluate the performance of compressed sensing for motion correction (CosMo) in compensating the respiratory motion of the heart in 3D late gadolinium enhancement (LGE) CMR.
Introduction
Late Gadolinium Enhancement (LGE) cardiovascular magnetic resonance (CMR) is the established gold standard for the non-invasive identification of myocardial scarring and fibrosis. However, the acquisition of high-quality 3D LGE datasets is frequently hampered by respiratory motion. Conventional techniques utilize diaphragmatic navigators to gate data acquisition to the end-expiratory phase. While this reduces motion artifacts, it is inherently inefficient; the scanner remains idle during a large portion of the respiratory cycle, often extending scan times by a factor of two or three. Such prolonged acquisitions can lead to gadolinium contrast washout, reducing the diagnostic signal-to-noise ratio, and pose challenges for patient compliance.
Methodology: The CosMo Technique
The researchers introduced CosMo (Compressed Sensing for Motion correction), a novel approach implemented on a 1.5T Philips CMR system designed to provide high-resolution 3D imaging within a fixed, short scan time. The technique optimizes the acquisition of "k-space"—the raw data domain of MRI—through a hybrid strategy:
- Hybrid Acquisition: The central 10% of k-space, which dictates the primary image contrast and signal, is acquired prospectively within a narrow 7 mm respiratory gating window.
- Fixed-Time Completion: The remaining 90% of k-space is acquired without respiratory gating, ensuring the scan concludes in approximately 8 minutes regardless of the patient's breathing pattern.
- Retrospective Motion Correction: Following the scan, any data in the outer 90% of k-space that was recorded outside the optimal respiratory window is discarded to eliminate motion-induced blurring.
- Compressed Sensing Reconstruction: To compensate for the resulting undersampled data (the "holes" left by discarded segments), advanced compressed sensing algorithms are employed to mathematically reconstruct a complete, high-resolution 3D volume.
Results
The performance of CosMo was evaluated in a study involving 10 healthy volunteers and 9 patients with suspected cardiac conditions.
- Resolution and Scan Time: While both conventional and CosMo protocols were limited to roughly 8 minutes, CosMo provided a significantly higher isotropic spatial resolution of 1.7 x 1.7 x 1.7 mm³, compared to the 4.0 mm thick slices typical of conventional LGE.
- Gating Efficiency: CosMo achieved a remarkable gating efficiency of 94%. In contrast, traditional navigator-gated methods often operate at 30-50% efficiency, meaning the scanner is productive for less than half the total scan duration.
- Image Quality: Expert readers identified 100% of CosMo-acquired images as diagnostic. On a 4-point scale for respiratory artifacts (where 4 represents no artifacts), CosMo scored 3.1 ± 0.7, demonstrating effective motion suppression.
Discussion
The transition to isotropic resolution (1.7 mm in all directions) represents a major clinical advantage. It allows clinicians to reformat and view the heart from any anatomical angle without the loss of detail associated with thick-slice imaging.
Furthermore, the high through-plane resolution significantly reduces "partial volume averaging." In standard 4.0 mm slices, small or thin areas of fibrosis can be averaged with adjacent healthy tissue, causing them to appear larger, fainter, or even disappear entirely. CosMo's thinner slices allow for more precise localization and characterization of scar morphology.
Conclusion
CosMo demonstrates the potential of compressed sensing to overcome the traditional trade-off between scan time and image resolution in cardiac MRI. By providing motion-corrected, high-resolution 3D LGE images in a predictable timeframe, it offers a pathway toward more efficient and accurate clinical diagnosis of myocardial disease. Future work will focus on validating these findings in larger clinical cohorts and optimizing reconstruction times for real-world clinical workflows.
Summary updated