The effect of acquisition duration on left ventricular volumes and function in free-running cardiac and respiratory motion-resolved 5D whole-heart cine MRI: A direct comparison between 0.55T and 1.5T

Background: Cardiac MRI has relied on 1.5T and 3T systems, but advances in high-performance low-field scanners offer lower cost, easier siting, and greater accessibility.¹ Yet, CMR remains complex and time-consuming. A free-running 5D framework can overcome these barriers by enabling high-resolution imaging without ECG gating, breath-holds, or view planning, improving efficiency and comfort.^2,3 Combining a free-running framework with high-performance low-field MRI may make cardiac imaging more accessible and patient-friendly, though its full potential remains unexplored. This study evaluates the impact of acquisition time on left-ventricular volume and function measurements in 5D free-running imaging at 0.55T, using standard 2D breath-hold cine at 1.5T as reference.

Methods: Sixteen healthy adults underwent standard 2D breath-hold cine CMR at 1.5T and, within 3 months, 5D free-running imaging at 0.55T with scan durations ranging from 1 to 9 min. Free-running 5D imaging used a continuous non-selective 3D radial bSSFP sequence with a spiral phyllotaxis readout, without ECG or respiratory gating (TE 2.7ms, TR 5.4ms, FOV 220 mm³, resolution 1.4mm³, FA 88º, BW 401 Hz/px). Data were reconstructed using an automated motion-resolved compressed-sensing framework to generate 5D images, which were reformatted into standard 2D cardiac views and analyzed for left-ventricular ejection fraction (EF), end-diastolic (EDV) and end-systolic volumes (ESV) and rated on image quality and segmentation confidence (using a qualitative 1 to 4 Likert scale). Agreement in volumes and function between techniques was evaluated and image quality and observer confidence were compared. Intra- and inter-observer agreement were also evaluated for all parameters.

Results: All 16 participants (median age 24 years, 50% female) successfully underwent CMR at both 1.5T and 0.55T (Fig. 1). No significant differences in EF were found between free-running imaging at 0.55T (for all scan durations) and 2D cine at 1.5T. Bland-Altman analysis showed small biases of -2.6% (1min), -1.2% (3 min), -1.8% (5 min), 0.3% (7 min), and 1.5% (9 min), with the narrowest limits of agreement at 7 min (-2.8% - 3.4%) (Fig. 2). EDV was significantly underestimated at 1 min (-20mL, p=0.024), 3 min (-11mL, p=0.020), 7 min (-9mL, p=0.049), and 9 min (-12mL, p=0.008). ESV showed no significant bias at 1 min (-4mL), 3 min (-3mL), 5 min (-2mL), and 7 min (-3mL), but was underestimated at 9 min ( -7mL, p=0.013). Image quality and segmentation confidence improved with increasing scan duration (mean scores 1.0/1.3 at 1 min, 1.4/1.6 at 3 min, 1.9/2.2 at 5 min, 1.9/2.3 at 7 min, and 2.3/2.5 at 9 min, respectively) but remained below 2D cine (4.0/4.0).

Conclusion: This study demonstrates the feasibility of 5D free-running CMR at 0.55T across a range of acquisition times, showing promise for robust functional assessment with simplified workflow. However, further optimization is needed to improve volumetric accuracy and image quality before broader clinical adoption.

Image examples acquired using reference standard 2D breath-hold cine imaging at 1.5T and using 5D free-running imaging with various durations (1, 3, 5, 7 and 9 min) at 0.55T. Various short-axis (from base to apex) and long-axis (both left-ventricular two-chamber [L2CH] and four-chamber [4CH]) cardiac views are shown, all in the end-diastolic phase of the cardiac cycle. Note that the 2D cine images have an 8 mm slice thickness compared to the 1.4 mm reformats of the free-running images.
Correlation plots and Bland–Altman analysis of the left-ventricular ejection fractions (in %) as derived from 5D free-running imaging using various durations (1, 3, 5, 7 and 9 min) at 0.55T compared to reference standard 2D breath-hold cine imaging at 1.5T. The solid red line and red shaded area in each Bland–Altman plot indicate the bias with its 95% confidence interval, while the black dotted lines indicate the limits of agreement. All axes indicate absolute percentages.