Purpose

Accurate and reproducible whole-slice and segmental circumferential strain assessed rapidly in three short-axis slices would provide valuable diagnostic and prognostic information and facilitate an efficient clinical workflow. For spiral cine DENSE, a well-established strain MRI method, three slices are typically acquired with each slice requiring one breathhold (14 heartbeats)^1,2,3. We have previously demonstrated multiband (MB) DENSE, acquiring three slices in two breathholds; however, slice separation artifacts and low signal-to-noise ratio presented challenges for further acceleration. Here we present improved results by modifying the MB phase modulation method and utilizing an improved iterative MB reconstruction model.

Methods

We implemented a modified spiral cine DENSE sequence with variable flip angle MB excitation, outer volume suppression³, and golden angle-rotation (Figure 1A). To acquire three slices in a single breathhold with temporal resolution of 30ms, the acquisition protocol uses 4 spiral interleaves³. However, when the number of interleaves is not a multiple of the MB factor, slice separation is challenging (for example, when MB=3 and number of interleaves = 4, as in Figure 2A). To overcome this challenge, we employed CAIPIRINHA phase modulation corresponding to MB=4 to simultaneously excite 3 slices using a 4 interleave acquisition (termed MB3+, Figure 1A). Experimentally, we compared CAIPIRINHA MB=3 phase modulation, Hadamard encoding phase modulation⁴ and MB3+. These methods were compared using a non-Cartesian slice GRAPPA (NCSG) reconstruction⁵. Single-band DENSE images at matched slice locations served as reference standards for comparison.
To improve the reconstruction, we developed a self-calibration method to fit NCSG kernels⁵ and coil sensitivity maps for each slice (Figure 1B). Specifically, the artifact-generating T1-relaxation echo was removed, followed by phase demodulation, NUFFT and temporal averaging of images to generate kernels and sensitivity maps without calibration scans. Next, the reconstruction employed a non-cartesian slice-low-rank approach based on a recently-developed Cartesian slice-L+S method⁶, which enforces MB data consistency, uses in- and through-plane coil information, exploits golden-angle rotation and enforces temporal low rank of multiple slices to reduce slice separation artifacts (Figure 1C). The slice-low-rank reconstructed images were compared to non-iterative NCSG reconstructed images⁵ and reference single-band DENSE images at matched slice locations. We evaluated the single-breathhold MB DENSE method in 4 human volunteers, where 3 slices with 2D displacement encoding were simultaneously acquired in a single breathhold (14 heartbeats). For one volunteer, datasets were acquired using three different phase modulation methods. MRI was performed on a 3T system (Prisma, Siemens) using 24-30 RF receiver channels.

Results

Figure 2A compares NCSG and slice-low-rank, using MB3 phase modulation in the acquisition. Slice-low-rank shows better image quality and is more similar to reference images than NCSG. Figure 2B illustrates the comparison of different MB phase modulation strategies, where NCSG-reconstructed images using MB3, Hadamard, MB3+ and reference single-band images are shown. MB3+ shows higher SNR of the myocardium than MB3 and Hadamard (arrows). Figure 3 illustrates example results from one volunteer, where MB3 phase modulation was used along with the NCSG reconstruction, and these images are compared to ones using MB3+ phase modulation and the slice-low-rank reconstruction. MB3+ combined with slice-low-rank shows significant improvement of image quality compared to MB3 with NCSG, demonstrating fewer artifacts and better SNR. Figure 4 illustrates example slice-low-rank and MB3+ results for 3 slices imaged simultaneously, and, adjacently, single-band reference images. The MB images show image quality similar to single-band images at basal, mid-ventricular and apical locations for both magnitude and phase. Example circumferential strain maps (Ecc) and segmental strain-time curves also show a close correspondence between the single-band and optimized MB methods (Figure 5A). Correlation and Bland-Altman plots in Figure 5B, and intraclass correlation of 0.98 quantitatively show excellent agreement between single-breathhold optimized MB DENSE and single-band DENSE for all segments and all three slices of four volunteers.

Discussion

The proposed MB spiral cine DENSE method with MB3+ phase modulation, self-calibration, golden-angle rotation, and a slice-low-rank reconstruction provides single-breathhold strain imaging of the left ventricle with excellent agreement compared to single-band cine DENSE strain. When combined with fully-automated deep-learning strain analysis⁷, these methods can provide accurate and reproducible global and segmental strain imaging with an efficient clinical workflow, requiring the addition of just a single breathhold to any 3T cardiac MRI protocol.

Acknowledgements

This work was supported by R01HL147104.

References

[1] Zhong X, Spottiswoode BS, Meyer CH, Kramer CM, Epstein FH. Imaging three-dimensional myocardial mechanics using navigator-gated volumetric spiral cine DENSE MRI. Magn Reson Med 2010;64(4):1089-1097. [2] Chen X, Yang Y, Cai X, Auger D, Meyer CH, Salerno M, Epstein FH. Accelerated two-dimensional cine DENSE cardiovascular magnetic resonance using compressed sensing and parallel imaging. J Cardiovasc Magn Reson 2016;18(1):38. [3] Tayal U, Wage R, Ferreira PF, Nielles-Vallespin S, Epstein FH, Auger D, Zhong X, Pennell DJ, Firmin DN, Scott AD, Prasad SK. The feasibility of a novel limited field of view spiral cine DENSE sequence to assess myocardial strain in dilated cardiomyopathy. Magnetic Resonance Materials in Physics, Biology and Medicine. 2019 Jun;32(3):317-29. [4] Yang Y, Meyer CH, Salerno M, Kramer MK, Epstein FH. Whole-heart spiral simultaneous multi-slice first-pass myocardial perfusion imaging. Magnetic Resonance in Medicine, 81, 852-862,2019. [5] Sun C, Yang Y, Cai X, Salerno M, Meyer CH, Weller D, Epstein FH. Non-Cartesian slice-GRAPPA and slice-SPIRiT reconstruction methods for multiband spiral cardiac MRI, Magnetic Resonance in Medicine, 83, 1235-1249, 2020. [6] Sun C, Robinson A, Schumann C, Weller D, Salerno M, Epstein FH. Multiband first-pass myocardial perfusion MRI using a slice-low-rank plus sparse model. ISMRM 28th Annual Meeting, Virtual meeting, 2020, p.1097. [7] Ghadimi S, Auger DA, Feng X, Sun C, Meyer CH, Bilchick KC, Cao JJ, Scott AD, Oshinski JN, Ennis DB, Epstein FH. Fully-automated global and segmental strain analysis of DENSE cardiovascular magnetic resonance using deep learning for segmentation and phase unwrapping. J Cardiovasc Magn Reson 2021;23(1):20.