Echo Planar Imaging Enables Single Breath-Hold 3D Late Gadolinium Enhancement

Background: Late Gadolinium Enhancement (LGE) imaging is the gold-standard technique for assessing myocardial viability [1, 2] but can suffer from long scan times. Clinically, LGE is performed either over a series of breath-holds, during which a set of 2D images are acquired; or in free breathing, where the acquisition can be either multi-slice 2D or fully 3D and breathing motion is compensated for using a navigator signal or data rejection and averaging [3, 4]. These methods all take 5–10minutes [5].

Recent work has proposed acquiring highly undersampled images to enable 3D images to be acquired in one breath-hold [6, 7]. However, either the required breath-holds were very long or compressed sensing with very high acceleration, which can result in residual blurring or aliasing artefacts. This study proposes using a fast 3D Echo-Planar Imaging (EPI) acquisition, allowing single breath-hold 3D LGE scans.

Methods: An ECG-triggered, inversion recovery–prepared, 3D EPI research sequence [8, 9] was implemented, playing one inversion pulse and a series of EPI shots in each R–R interval (figure 1). K-space was undersampled with a CAIPIRINHA pattern and segmented into skipped-CAIPI shots [10–12]. The shots were divided based on the region of k-space they sampled and one shot from each region was played in each R–R. Within an R–R, shots were acquired in a centric order, with the shots from the central region acquired first, unless the required TI and acquisition window could not be realised with this ordering, in which case a linear order was used.

Images were reconstructed using volumetric GRAPPA [13] with an untriggered, FLASH reference scan used for auto-calibration; and a three-line navigator signal for phase correction. The reference scans were performed at the start of the scan.

Images with the proposed sequence were acquired in 25 patients referred for clinical CMR exams on a 3T MRI system (MAGNETOM Vida, Siemens Healthineers AG, Forchheim, Germany) following IRB approval and informed consent and compared to the clinical 2D breath-hold images. The proposed sequence was acquired after the clinical exam was complete with the following parameters: transverse orientation; 320×264×120mm³ FOV; 144×120×56 matrix; 2.2mm isotropic resolution; 2×2 undersampling; echo train length 12; binomial water excitation pulses; flip angle 20^◦; TE 6.4ms; echo spacing 0.63ms; bandwidth 2042Hz/px. The inversion time was chosen to null viable myocardium and the number of echo trains played per R–R interval was chosen to fill the diastolic still window determined from a 4-chamber cine.

Results: The breath-hold times for the proposed sequence were 17.2 ± 2.0s. All scans were successfully completed, and visual inspection confirmed similar enhancement patterns between the proposed and clinical scans (figure 2).

Conclusion: The proposed sequence was able to acquire 3D LGE images with feasible breath-hold times, with the isotropic resolution enabling reformatting and LGE visualisation in the desired views (figure 3).

Figure 1: A diagram showing the structure of the proposed sequence. Three reference scans—a noise scan, GRE reference scan and three-line navigator—are played before a dummy heartbeat and the segmented image acquisition (shown top). During each heartbeat of the image scan an inversion pulse is played followed by a series of EPI echo trains (shown bottom).
Figure 2: Images showing a comparison between the clinical (left) and proposed (right) scans. One clinical short axis slice and a short axis reformat of the proposed scan are shown top and a similar 3-chamber pair is shown bottom. Both scans show similar patterns of lateral subendocardial enhancement.
Figure 3: Images showing multiple reformats of an image acquired using the proposed sequence. 4, 3, 2-chamber and short axis views are shown. Extensive LGE can be observed.