Improving Temporal Resolution of Single-Shot Late Gadolinium Enhancement Imaging via Partial Fourier Sampling and DL Reconstruction

Background: Single-shot phase-sensitive late gadolinium enhancement (ssh PS LGE) imaging has gained widespread clinical adoption due to its rapid acquisition time compared to conventional segmented PS LGE^1,2. However, since the entire dataset is acquired within a single heartbeat, the spatiotemporal resolution is inherently limited. According to SCMR recommendations³, the spatial resolution should range from 1.4 mm to 1.8 mm, and the temporal resolution should be below 200 ms, which are difficult to achieve with conventional ssh PS LGE protocols, even with parallel imaging. In this study, we propose an accelerated ssh PS LGE acquisition strategy by incorporating partial Fourier sampling in both frequency and phase encoding directions. Combined with AIR Recon DL and homodyne reconstruction, we are aiming to substantially enhance the temporal resolution of ssh PS LGE while preserving high spatial resolution.

Methods: Post-contrast ssh PS LGE data was acquired from 25 patients (age = 52.5 ± 21.1, 16 males) in both short-axis (SA) and 4-chamber (4CH) views using a clinical 1.5 T MRI scanner (GEHC). For comparison, back-to-back LGE scans were performed using both the conventional and enhanced protocols, as detailed in Table 1. In the enhanced ssh PS LGE protocol, partial Fourier sampling was applied in both the frequency and phase encoding directions, enabling shorter TE, TR, and improved temporal resolution relative to the conventional protocol. The missing k-space data was reconstructed using homodyne reconstruction and AIR Recon DL. An experienced cardiac radiologist quantitatively evaluated the image quality from both protocols using a 5-point Likert scale. The Wilcoxon signed-rank test was used to evaluate differences in image quality between the two protocols, with statistical significance defined as a P-value less than 0.05.

Results: The enhanced ssh PS LGE protocol demonstrates better image quality with increased sharpness and decreased blur compared to the conventional protocol (Figures 1, 2). No artifacts were observed with the enhanced protocol, despite the use of partial Fourier sampling. In the comparison of image quality (Figure 3), the enhanced protocol yielded significantly higher mean scores than the conventional protocol on both imaging planes. On the SA plane, the mean score was 4.08 ± 0.49 versus 3.76 ± 0.52 (p < 0.05). In the 4CH plane, the scores were 4.16 ± 0.37 versus 3.76 ± 0.52 (p < 0.01). Overall, image quality improved in 40% of the cases.

Conclusion: In this study, we proposed the use of partial Fourier sampling combined with Homodyne reconstruction and AIR Recon DL to enhance the temporal resolution of ssh PS LGE imaging, while preserving the high spatial resolution recommended by SCMR. The enhanced protocol led to notable improvements in image quality—particularly in image sharpness—which significantly improved scar visualization and diagnostic confidence. Importantly, no artifacts attributable to partial Fourier sampling or Homodyne reconstruction were observed.

Example of the conventional and enhanced ssh PS LGE protocol with partial TE and partial NEX enabled.

	Conventional protocol	Enhanced protocol
FOV [cm]	36.0
Phase FOV	0.9	0.9
Slice Thickness [mm]	8.0
Flip angle [°]	45
Bandwidth [Hz]	83.33
Acceleration	ARC x 3
NEX	1	0.9
TR [ms]	3.6	3.1
TE [ms]	1.5	1.1
Spatial Res [mm]	1.8	1.8
Temp res [ms]	258	203

 Figure 1. Representative ssh PS LGE images from a 19-year-old male patient acquired during the 6-month follow-up. The patient presented to the emergency department with one week history of cough, fever, and chest pain. His cardiac MR findings were consistent with acute myocarditis including myocardial edema with elevated T2 value, elevated T1/ECV value, and extensive delayed enhancement. Patient returned 6 months later for follow up with resolution of parametric mapping abnormalities and persistent delayed enhancement. Ssh PS LGE images acquired with the enhanced protocol (A, C, E) demonstrate improved image sharpness and decreased image blur compared to those acquired with the conventional protocol (B, D, F), as highlighted by the yellow arrows.
Figure 2. 83-year-old male patient with history of sarcoid referred for evaluation of non-sustained ventricular tachycardia. Ssh PS LGE acquired with the enhanced protocol (A) demonstrates improved image sharpness with better visualization of the inferolateral wall patchy midmyocardial delayed enhancement (yellow arrow) over conventional (C). Additional short axis image demonstrating patchy mid myocardial delayed enhancement within the mid anterolateral wall (white arrow) with improved image sharpness and less image blurring with the enhanced ssh PS LGE (B) over conventional (D).
Figure 3. Comparison of image quality scores for ssh PS LGE acquired using conventional and enhanced protocols (n = 25). On the short-axis (SA) plane (left), the mean score was 4.08 ± 0.49 for the enhanced protocol versus 3.76 ± 0.52 for the conventional protocol (p < 0.05). On the 4-chamber (4CH) plane (right), scores were 4.16 ± 0.37 versus 3.76 ± 0.52, respectively (p < 0.01).