Functional cardiac MRI scans use ECG signals for triggering acquisition but at higher fields and with demands from, e.g., fetal imaging, new triggering methods are sought¹. We propose feasibility of triggering to carotid ultrasound using an MRI compatible probe, moving towards full ultrasound-controlled hybrid cardiovascular MRI, using directly the mechanical flow signal or tissue displacement, instead of electrical stimulation. Unlike previous studies^2,3 we used spatially-resolved Doppler imaging. Retrospective processing using Metric Optimized Gating (MOG)^4-6, is included for comparison.

ECG (gold-standard), Doppler and MOG triggering were compared in 5 healthy volunteers (age 26±5years, weight 75±11kg, 3 female 2 male). The 3T MRI protocol consisted of 2- and 4-chamber (2C, 4C) and short axis (SA) cines (TE/TR 1.4/39.2ms, resolution 1.63x1.63x6mm, retrospective gating, 25 phases) and aortic phase contrast (PC) flow (TE/TR 2.5/37.1ms, resolution 1.98x1.98x6mm, retrospective gating, 30 phases). Typical breath-hold duration was 9RR for cine and 24 for PC flow.

Intraoperatory real-time US imaging was performed simultaneously using an MRI compatible transducer (128 element, bandwidth 5-10MHz, pitch 0.15mm, frame rate 20-30fps, shielded external casing dimension 2x3.5cm) designed to minimize susceptibility artifact, RF- and gradient-switching interferences. Pulse-colour Doppler was analysed by an external computer, generating a trigger from a time adaptive threshold of cumulative signal inside a user-defined ROI ensuring robustness to motion. The transducer was positioned in bore using an Innomotion robotized arm or elastic belt, targeting the left common carotid artery. Figure 1 illustrates the US set up and images obtained.

‘Fake trigger’ images used simulated RR 20-30% longer than reality to compare as a poorly gated image and for correction using the MOG (retrospectively corrected k-space).

Images were assessed for quality and functional parameters. Edge gradient sharpness was quantified on septal regions throughout the cycle on Modulus/Laplacian/Median image transforms. Time-resolved flow curves were cross-compared between triggering methods. Left ventricular (LV) volume, ejection volume and flow velocity were calculated using semi-automatic segmentation and manual adjustment (Osirix).

MR imaging was completely free of any potential interferences generated by US. Detectable but non-significant artifacts were observed on colour Doppler during MR acquisition but signal conversion was optimized to be unaffected. Maintaining placement of the probe was successful and ‘patient-friendly’ in all cases.

Figure 2a shows 4C and SA cine frames showing image quality. High-resolution cine was also possible and is shown along with normal resolution for Doppler- and ECG-triggered (figure 2b).

Figure 3a shows Modulus/Laplacian/Median filtered images. Quantification of the septum for SA and 4C for 5 subjects showed 5 cases had no significant difference ECG:Doppler (0.339<p<1), 3 with Doppler better (0<p<0.0015) and 2 with ECG better (0<p<0.005).

Plotting LV surface over the cycle (SA and 4C) gave no difference between ECG and Doppler (figure 3b). Fake-trigger gave incorrect function and MOG succeeded in recuperating LV quantification. Mean difference between surfaces derived from the different triggered acquisitions (all subjects and phases) was 1.67 ECG:Doppler, 3.73 Fake:ECG and 3.28 Fake:Doppler with ICC>0.85 for ECG, Doppler and MOG.

Ejection volumes (4C, figure 3c) were not different between methods, but showed larger dispersion for Fake (in 2 cases Fake images not significantly degraded). ECG and Doppler were significantly different to Fake (p<0.04 without uncompromised Fake).

ECG and Doppler triggered flow showed a higher peak velocity than Fake triggered images and in some cases fake showed several peaks per cycle (figure 4a). The normalised peak flow in the ascending and descending aorta was not different between ECG and Doppler (p=0.475). The time displacement showed, for example, Doppler trigger 600ms later than ECG in an RR of 900ms.

Doppler triggering is based on hydrodynamic phenomena, which are related to the PC flow-encoding mechanism unlike cardiac electrical stimulation.

We observe directly major arteries (common carotid) with similar functional properties to the aorta, not small peripheral vessels with different properties used for PPU.

Cross correlation between velocity measurements by US Doppler and PC MR become feasible under conditions of sufficient acoustic window to the respective blood vessel.

Our method does not require direct US imaging of the heart itself which is a complicated acoustic problem inside a closed bore MRI.

PC flow and cine images were successfully obtained in healthy volunteers with ECG-, Doppler- triggering and MOG. Image quality is highly comparable and accurate functional parameters accessible. Quantitative images are obtained in the absence of an ECG signal with Doppler fast enough to trigger cardiac function images.

The hardware platform is designed to further enable advanced cardiac imaging using real-time slice tracking locking imaging planes to respiratory and/or cardiac motion under free breathing, allowing increased patient comfort and longer scantimes.