Flow-robust and sampling-efficient free-running 5D self-gated cardiac MRI at low-field

Background: Plug-and-play low-field cardiac MRI (CMR) is becoming a reality[1] that could redesign the accessibility of CMR worldwide with reduced cost and increased accessibility for obese or claustrophobic patients thanks to the larger bore size. However, several challenges remain to be overcome for a streamlined CMR workflow, including robust motion management, prolonged acquisition duration and adequate contrast between blood and cardiovascular structures. Recent free-running framework (FRF)[2] implementations at low-field[3] have started to address these challenges but remain to be perfected. A lung-dedicated sequence, bStar[4], provides an improved k-space trajectory for acquisition efficiency and robust self-gating, albeit with flow-sensitive bipolar gradients. Hence, we evaluated a combined FRF bStar and a proposed flow-robust extension to bStar, termed bNovae.

Methods: A numerical simulation including pulsatile blood flow was implemented to compare FRF signal behavior between 1^st-moment M1-compensated full-line and flow-sensitive bStar. Our extension to the bStar module, bNovae, extends the 2^nd echo from 50% to 85% of a full-line, thus reducing M1 to decrease flow-sensitivity. Sampling duty cycle was reported for each sequence. Conventional full-line, bStar and bNovae FRF were implemented in-house on a commercial 0.55T MRI with parameters described in Table 1. After trajectory GIRF correction[5], established FRF self-gating[6] and Total Variation Compressed Sensing[2] reconstruction provided 5D complex images. Three volunteers (3 men, 32±7 y.o.) were imaged. Aortic signal variations were analyzed in simulation and in experimentation. Images were evaluated for contrast ratio and sharpness between left ventricular blood pool and myocardium.

Results: Numerical simulation (Fig.1) revealed on average 69% signal loss using bStar due to pulsatile aortic flow compared to full-line free-running. The extension of the 2^nd echo to 85% reduced this to a simulated average 20% signal loss while increasing sampling duty cycle by almost 50% compared to full-line FRF (Fig.1 & Table 1). Experimentally, compared to full-line, average signal loss was 47% for bStar and 3% for bNovae. Improved blood-to-myocardium contrast (Fig.2) was confirmed in vivo with bNovae CR approaching full-line CR (Table 1), and with more stability over the cardiac cycle. bStar CR suffered greatly from velocity-dependent blood signal loss while full-line bSSFP showed dark jets during early systole. However, bStar blood-to-myocardium sharpness was the highest, followed by that of bNovae. Full-line images had the lowest sharpness matching visually observed image blur (Fig. 2).

Conclusion: The combination of whole-heart FRF with motion-optimized high-sampling efficiency bNovae is well-suited for plug-and-play low-field CMR. bNovae overcomes motion-induced signal loss despite a constrained gradient system and improves low-field signal. Further investigation is warranted to also leverage bNovae’s two echoes for fat-water separation.

Comparison of Free-Running MRI Acquisition Strategies for Cardiac Imaging

 

	full-line FRF	bStar FRF	bNovae FRF
Acquisition Parameters
TE/TR (ms)	2.7 / 5.4	0.4, 2.8 / 3.3	0.4, 3.7 / 5.1
Flip angle (°)	80
FOV (mm³)	300 × 300 × 300
Spatial resolution (mm³)	1.5 × 1.5 × 1.5
Sampling Strategy
Number of segments	25
Number of interleaves	2,812	4,601	2,978
Total acquisition time (min)	6:00
Reconstructed respiratory bins × cardiac phases	4 × 20
Performance Metrics
Sampling duty cycle† (%)	44	73	64
Blood-myocardium contrast ratio‡ (%)	90 ± 31	31 ± 18	72 ± 19
Image sharpness (a.u.)	0.61 ± 0.29	2.68 ± 3.40	1.12 ± 1.95

---

Abbreviations: FRF, free-running framework; FOV, field of view; TE, echo time; TR, repetition time; a.u., arbitrary units.

Notes: † Sampling duty cycle calculated as ADC duration/TR × 100% ‡ Blood-myocardium contrast ratio calculated as (B-M)/M × 100%, where B represents blood signal intensity and M represents myocardial signal intensity

 Free-running sequence modules (A), with ADC sampling in gray, were included in a numerical simulation (B) to confirm bStar high flow sensitivity (C). By extending the 2nd echo, in proposed bNovae, flow-induced signal loss was greatly reduced while maintaining an ultra-short echo time and a high sampling duty cycle.
A: Measurements of free-running (FRF) signal variations in the aortic root confirmed bStar signal loss, as anticipated by numerical simulations. B: Coronal views of key frames within the cardiac cycle showed severe signal loss in the aortic root: from intravoxel dephasing in full-line FRF (purple arrow) and from flow-induced accumulated phase in bStar FRF (red arrow). Thanks to increased sampling efficiency, bStar and bNovae images were also sharper than full-line images that suffered from a global mild blur.