How low can we go? The effect of acquisition duration on cardiac volume and function measurements in free-running cardiac and respiratory motion-resolved five-dimensional whole-heart cine magnetic resonance imaging at 1.5T

Abstract

Background: Cardiovascular magnetic resonance (CMR) is the gold standard for assessing cardiac volumes and function using two-dimensional (2D) breath-held cine imaging. This technique, however, requires a reliable electrocardiogram (ECG) signal, repetitive breath-holds, and the time-consuming and proficiency-demanding planning of cardiac views. Recently, a free-running framework has been developed for cardiac and respiratory motion-resolved five-dimensional (5D) whole-heart imaging without the need for an ECG signal, repetitive breath-holds, and meticulous plan scanning. In this study, we investigate the impact of acquisition time on cardiac volumetric and functional measurements, when using free-running imaging, compared to reference standard 2D cine imaging.

Methods: Sixteen healthy adult volunteers underwent CMR at 1.5T, including standard 2D breath-held cine imaging and free-running imaging using acquisition durations ranging from 1 to 6 min in randomized order. All datasets were anonymized and analyzed for left-ventricular end-systolic volume (ESV) and end-diastolic volume (EDV), as well as ejection fraction (EF). In a subset of data, intra- and inter-observer agreement was assessed. In addition, image quality and observer confidence were scored using a 4-point Likert scale. Finally, acquisition efficiency was reported for both imaging techniques, which was defined as the time required for data sampling divided by the total scan time.

Results: No significant differences in left-ventricular EDV and ESV were found between free-running imaging for 1, 2, 3, 5, and 6 min and standard 2D breath-held cine imaging. Biases in EDV ranged from -2.4 to -7.4 mL, while biases in ESV ranged from -3.8 to 2.1 mL. No significant differences in EF were found between free-running imaging of any acquisition duration and standard 2D breath-held cine imaging. Biases in EF ranged from -2.8% to 0.94%. Both image quality and observer confidence in free-running imaging improved when the acquisition duration increased. However, they were always lower than standard 2D breath-held cine imaging. Acquisition efficiency improved from 13% for standard 2D cine imaging to 50% or higher for free-running imaging.

Conclusion: Free-running CMR with an acquisition duration as short as 1min can provide left-ventricular cardiac volumes and EF comparable to standard 2D breath-held cine imaging, albeit at the expense of both image quality and observer confidence.

Keywords: 5D; CMR; Cardiac MRI; Free-breathing; Free-running; Self-gating.

Graphical abstract

Fig. 1

Schematic overview of the free-running framework. This framework consists of a 1) continuous 3D radial acquisition with a spiral phyllotaxis readout, 2) self-gating signal extraction using principal component analysis and SOBI, 3) data sorting and binning using the cardiac and respiratory self-gating signals, 4) CS reconstruction, and 5) data reformatting using the standard clinical 2D cine geometries. SOBI second-order blind identification, SG self-gated, SI superior-inferior, CS compressed sensing, 2D two-dimensional, 3D, three-dimensional

Fig. 2

Image examples acquired using reference standard 2D breath-held cine imaging and as derived from the 3D free-running acquisitions using different scan durations (1 to 6 min). Various short-axis (from base to apex) and long-axis (both left-ventricular two-chamber and four-chamber) 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 cine images. 2D two-dimensional, 3D, three-dimensional

Fig. 3

Correlation plots and Bland–Altman analysis of the left-ventricular ejection fractions as derived from free-running cine using various durations (1–6 min) compared to reference standard 2D breath-held cine imaging. 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. 2D two-dimensional

Fig. 4

Bar plots indicating the Likert scale score frequencies for both overall image quality (upper panel) and observer confidence in segmentation (lower panel) for free-running cine imaging using various acquisition durations (1 to 6 min) and reference standard 2D cine imaging. For each type of acquisition, the average score is indicated. The asterisks indicate a significant difference compared to reference standard 2D cine imaging. 2D two-dimensional