Spiral cine displacement encoding with stimulated echoes (DENSE) MRI at 0.55 T

Background: Lower-field MRI applications are emerging due to reduced system costs, simplified siting requirements, improved patient comfort, and increased patient access^1-3. However, at 0.55T, dedicated development and optimization of sequences and imaging protocols are necessary, particularly for cardiovascular applicaitons⁴. Displacement encoding with stimulated echoes (DENSE) MRI is essential for accurate myocardial strain and cardiac function evaluation^5-7. To our knowledge, DENSE MRI has not been performed and validated in low fields (< 1.5T). We aim to investigate the feasibility of spiral cine DENSE MRI on a clinically available 0.55T scanner with dedicated sequence protocol and image denoising, and compare with its performance at 3T.

Methods: Three healthy subjects (3 male, age: 32±7 years) were scanned on clinical 3T (Prisma, Siemens) and 0.55T scanners (Free.Max, Siemens) within one week. Breath-hold 2D cine DENSE was acquired on a short-axis mid-level slice^6,7. Imaging parameters at 3T included FOV=360×360mm², resolution=2.8×2.8mm², slice thickness=8mm, TR/TE=16/1.08ms, cardiac phases=22-23, spiral interleaves per image=6, displacement encoding frequency=0.10cycles/mm, through-plane dephasing frequency=0.08cycles/mm. For the 0.55T protocol, selected parameters were adjusted to improve SNR and adapt to the longer fat-suppression pulse, including FOV=410×410mm², resolution=3.2×3.2mm², slice thickness=10mm, TR/TE=24/1.49ms, cardiac phases=15. To further improve SNR, a recently developed locally low-rank denoising technique⁸ leveraging the high-dimensional data (multiple displacement and phase-cycling encodings, coils and cardiac phases) was applied on 0.55T DENSE images. Both denoised 0.55T and non-denoised 3T DENSE images were processed^7,9 to generate displacement maps, circumferential strain (E_cc) maps, and strain-time curves in 6 short-axis left ventricular segments¹⁰. Bland-Altman analysis was performed to compare E_cc.

Results: Denoising effectively suppressed the noise and provided higher SNR in magnitude and phase images (Figure 1). The denoised 0.55T DENSE images provided comparable image quality as non-denoised 3T DENSE images. Results from denoised 0.55T and non-denoised 3T DENSE data showed similar trends and peak values in strain-time curves for 0.55T versus 3T (Figure 2). Bland-Altman analysis showed that the E_cc mean difference between denoised 0.55T and non-denoised 3T was 0.0196 (Figure 3). Our preliminary work demonstrated the feasibility of spiral cine DENSE MRI at 0.55T with adjusted sequence parameters and denoising.

Conclusion: We demonstrated the feasibility of spiral cine DENSE MRI at 0.55T and our preliminary results showed the E_cc strain measurement is comparable with that measured at 3T. Future work will further optimize the DENSE protocol at 0.55T that balances the image resolution, scan time and image SNR. A larger cohort of healthy and/or clinical subjects will be scanned in the near future to support comprehensive validation.

Figure 1. Example DENSE magnitude and phase images from two subjects.
Figure 2. Example Displacement map, circumferential Ecc strain map and strain-time curve from two subjects.
Figure 3. Bland-Altman plots comparing circumferential (Ecc) strain from denoised 0.55T DENSE versus non-denoised 3T DENSE. Because the TRs are different at 0.55T and 3T, only strain values at similar time points were compared. MD: mean difference. LoA: limits of agreement.