Introduction

Accurate left ventricular (LV) volumes play an important role in the management of cardiovascular disease. Standard cardiac magnetic resonance (CMR) with a retrospective electrocardiogram (ECG)-gated breath-hold cine is the reference standard for assessing LV volumes¹. This method segments the k-space for each cardiac phase across multiple heartbeats and assumes that all these heartbeats are the same duration. However, this technique is not suitable when there are excessive beat-to-beat variations such as arrhythmia or when patients are unable to hold their breath².

For these patients a prospectively ECG-triggered real-time (RT) technique is often used³, where the k-space for each cardiac phase is acquired in a single heartbeat, normally at the expense of spatial and temporal resolution⁴. Two drawbacks of ECG-triggered RT are that (1) the entire cardiac cycle is not covered, meaning end-diastolic volume (EDV) and ejection fraction (EF) may be underestimated⁵, and (2) although each slice is triggered from an R wave, the beat-to-beat variations can mean that the frame at which end-systole is captured can differ from slice to slice. This makes automated volumetric analysis impossible and manual analysis challenging.

A potential solution is a “retro-gated” RT sequence, where at each slice an R wave triggers the acquisition lasting 4 seconds to acquire 120 frames. This acquisition window covers ~4 heart beats, given a typical heart rate of 60bpm. The most consistent heart beat is then chosen across all slices by computing the median R-R interval for all acquired heart beats and selecting for each beat the best closest to mean, while excluding beats preceded by short beats. The images acquired in this best heart beat are then temporally interpolated to 30 output phases during Gadgetron reconstruction resulting in the “retro-gated” real-time output covering the entire cardiac cycle for each slice.

The aim of this study was to investigate the accuracy of the LV volume measurements derived from this retro-gated RT sequence in a cohort of healthy volunteers in sinus rhythm compared to the reference standard retrospective ECG-gated breath-hold sequence.

Method

CMR studies were performed on 24 healthy volunteers (mean age: 50y +/-14y) using a 1.5T scanner (MAGNETOM Aera, Siemens Healthineers, Erlangen, Germany) and the 60-channel cardiac coil. A left ventricular short-axis cine stack (LVSA) using a 2D balanced steady-state free precession (bSSFP) was acquired on each subject by three methods: (1) the reference breath-hold retrospectively-gated segmented sequence (bSSFP_SBH), (2) a standard prospectively triggered breath-hold real-time sequence (bSSFP_RTP) and (3) the “retro-gated” prospectively triggered free-breathing real-time sequence (bSSFP_RTRG). Sequence parameters are summarised in Figure 1.

Quantitative analysis of EDV, end-systolic volume (ESV), stroke volume (SV) and EF was performed by an experienced CMR technologist using CVI42 (v5.13.5, Circle Cardiovascular Imaging, Calgary, Canada). For each of the three LVSA sequences the epicardial and endocardial contours were automatically traced and manually corrected if necessary. The end-systolic and end-diastolic phases were automatically detected, based on the smallest and largest LV volumes over the entire cardiac cycle (Figure_2).

The quantitative results were compared using a linear mixed model ANOVA with post-hoc pairwise comparison. A p-value of less than 0.05 was considered statistically significant. Correlation and Bland-Altman analysis was performed to compare the quantitative measures for each sequence.

Results

All 24 participants had a regular sinus rhythm with the mean heart rate across all participants 63 ± 9 bpm. There was no statistically significant difference between the bSSFP_SBH, bSSFP_RTP and bSSFP_RTRG for all LV volume measurements (Figure_3).

There was a significant correlation between the bSSFP_RTRG and bSSFP_SBH for EDV (R2=0.87, p<0.05), ESV (R2=0.78, p<0.05), SV (R2=0.84, p<0.05) and EF (R2=0.56, p<0.05). There was also a significant correlation between the bSSFP_RTP and bSSFP_SBH for EDV (R2=0.89, p<0.05), ESV (R2=0.84, p<0.05), SV (R2=0.74, p<0.05) and EF (R2=0.40, p<0.05) (Figure_4).

Bland-Altman analysis (Figure_5) revealed a mean bias between the bSSFP_RTRG and bSSFP_SBH for EDV of 2.2mls (95% LoA, -20.8mls to 25.3mls), for ESV of 2.5mls (-11.9mls to 16.9mls), for SV of -0.3mls (-15.5mls to 14.9mls) and for EF of -1.2% (-8.3% to 5.9%). The mean bias between the bSSFP_RTP and bSSFP_SBH for EDV was -5.5mls (95% LoA, -25.7mls to 14.8mls), for ESV was -0.5mls (-12.5mls to 11.5mls), for SV was –4.9mls (-24.0mls to 14.2mls) and for EF was -1.1% (-9.8% to 7.7%).

Discussion

The bSSFP_RTRG showed high agreement for the volumetric analysis of the LV when compared to the current reference standard multi-breath-hold cine. As this technique is able to image the entire cardiac cycle it did not underestimate the EDV and EF as traditional bSSFP_RTP did (mean bias 2.2ms vs -5.5ms).

The consistent number of reconstructed cardiac phases in the bSSFP_RTRG ensured that analysis was straightforward as the end-systolic frame was the same for all slices in the LVSA. Although the spatial resolution of both the bSSFP_RTRG and bSSF_PRTP was less than that of the bSSFP_SBH it did not hinder the automated delineation of endo- and epicardial borders nor the automatic detection of end-diastole and end-systole.

Conclusion

Retro-gated RT CMR provides an accurate method for volumetric analysis of the LV and has the potential to provide a viable alternative to the standard retrospective-gated breath-hold method in patients who are unable to hold their breath or who suffer from arrhythmia.

Acknowledgements

No acknowledgement found.