Introduction

Myocardial T1 and T2 mapping are used to assess focal and diffuse cardiac pathologies. Cardiac cine MRI is also the gold standard for assessing cardiac motion and function. Single-acquisition 3D joint T1/T2 mapping with simultaneous cine motion resolution would offer significant efficiency benefits over standard gated, 2D, contrast-specific, serial imaging acquisition. Current free-breathing myocardial multi-parametric mapping techniques^1-3 have been reported at 1.5T, and still require ECG triggering or retrospective gating. MR Multitasking has demonstrated T1/T2 cine mapping without ECG gating and breath-holding, but only in 2D slices thus far⁴. We propose a 3D free-breathing, non-ECG, T1 and T2 cine mapping technique. The non-uniformity of the B1+ transmit field often causes a different flip angle to be experienced by tissue, so our 3D Multitasking framework additionally incorporates a novel B1+ mapping component⁵ to allow for accurate T1 mapping.

Methods

Sequence Design: The proposed sequence uses a prototype 3D stack-of-stars trajectory modified to collect low-rank tensor auxiliary data interleaved with image data. The image data is incremented by a tiny golden angle (i.e., 23.67°). The auxiliary data is acquired every 8th readout. Interleaved IR and T2-IR pulses are employed to generate the T1 and T2 contrasts (i.e., Fig. 1). FLASH excitation alternates between 3° and 10° for each recovery period to allow for B1+ robust T1 mapping^5-7.
Image Reconstruction: The images acquired in the Multitasking framework can be represented as a 6-way tensor $$$ \boldsymbol{\mathcal{A}} $$$ with voxel location index $$$ \mathbf{r}=(x,y,z) $$$, a T1 recovery dimension, a T2 decay dimension, a cardiac motion dimension, a respiratory motion dimension, and a flip angle dimension. The high correlation along and across multiple time dimensions induces $$$ \boldsymbol{\mathcal{A}} $$$ to be low-rank, which can be factorized as $$$ \boldsymbol{\mathcal{A}}_{(1)}=\mathbf{U} \mathbf{\Phi} $$$, where $$$ \boldsymbol{\mathcal{A}}_{(1)} $$$ is the mode-1 matricization of $$$ \boldsymbol{\mathcal{A}} $$$, $$$ \mathbf{U} $$$ contains spatial coefficient maps, and $$$ \mathbf{\Phi} $$$ contains matricized, separable multi-dimensional temporal basis functions describing T1/T2 relaxation, cardiac motion, respiratory motion, and FLASH readout flip angles. In reconstruction, $$$ \mathbf{\Phi} $$$ was first determined from the HOSVD of the low-rank tensor–completed auxiliary data. $$$ \mathbf{U} $$$ was then recovered by fitting $$$ \mathbf{\Phi} $$$ to the acquired imaging data $$$ \mathbf{d} $$$, according to the following optimization problem:$$ \mathbf{\hat{U}}=\text{arg}\underset{\mathbf{U}}{\operatorname{min}} ||\mathbf{d}-\Omega(\mathbf{FSU\Phi})||_2^2 + R(\mathbf{U}), $$ where $$$ \Omega $$$ is the undersampling operator, $$$ \mathbf{F} $$$ is the nonuniform Fourier transform operator, $$$ \mathbf{S} $$$ is the coil sensitivity operator, and $$$ R(\cdot) $$$ is a regularization functional promoting transform sparsity (in this case, wavelet sparsity of $$$ \mathbf{U} $$$).
T1,T2 Quantification: The native T1, T2 measurements are estimated from the signal model $$ s(A,\beta,B,T_1,T_2)=A \frac{1-e^{-TR/T_1}}{1-e^{-TR/T_1}cos(\beta \alpha)} [1+(Be^{-\tau/T_2}-1)(e^{-TR/T_1}cos(\beta \alpha))^n]\cdot sin(\alpha), $$ with amplitude factor $$$ A $$$, IR/T2-IR pulse efficiency $$$ B $$$, FLASH readout interval $$$ TR $$$, prescribed flip angle $$$\alpha$$$, Look-Locker B1+ factor⁵ $$$\beta$$$ and recovery time point $$$ n=1,2,..., N$$$, ( $$$ N $$$ is the number of FLASH pulses in a recovery period).
Data Collection: A 16-vial ISMRM NIST phantom and 7 human volunteers (3 male; 4 females) were imaged on a 3T scanner (MAGNETOM Vida, Siemens). Multitasking pulse sequence parameters were: TR/TE=3.6 ms/1.7 ms, recovery period = 2.5 s, spatial resolution=1.4x1.4x8.0 mm³, scan time=9min14s, 20 cardiac bins, 6 respiratory bins. Reference 2D T1 maps with MOLLI and 2D T2 maps with T2-prep FLASH (matched resolution, 1.4x1.4x8.0 mm³) were acquired at both systole and diastole during end-expiration breath-holds at basal, mid, and apical short-axis slices. A full short-axis stack of ECG-gated, breath-held, 2D bSSFP cine scans (1.3x1.3x8.0 mm³, TA=11s per slice) were acquired.

Results

T1 and T2 mapping results from the phantom are shown in Fig. 2. Diastolic and systolic T1, T2, and B1+ mapping results from one subject are shown in Fig. 3, including the basal, mid, and apical short-axis slices from the 3D multitasking sequence and reference 2D MOLLI and T2-prep FLASH sequences. Fig. 4 depicts the summary statistics for segment-wise T1/T2 values in human volunteers. Multitasking T1 maps reported higher myocardial T1 values (1495±47.8ms) than the MOLLI T1 values (1223±16.1ms). Multitasking T2 maps also produced slightly higher myocardial T2 values (43.8±1.3 ms) than the T2-prep FLASH T2 values (39.8±1.3 ms). Fig. 4 also shows the black-blood cine maps from the Multitasking output and the bright-blood cine maps from the bSSFP cine sequence.

Discussion

In phantom experiments, Multitasking measurements produced slightly higher ICC against gold standard IR-SE and SE measurements than the MOLLI and T2-prep FLASH sequences did, but all ICCs were >0.96. Although Multitasking produced substantially higher T1 measurements than MOLLI, MOLLI is known to underestimate T1; Multitasking measurements were closer to the reported values from SASHA at 3T⁸, which is known to be more accurate than MOLLI. Look-Locker B1+ maps showed a lower Look-Locker effect in the blood pool than myocardium, consistent with inflow.

Conclusion

3D MR Multitasking is feasible for myocardial T1/T2 mapping and cine imaging in the myocardium at 3T without ECG and breath-holds. Future studies will focus on improving the SNR of T1-T2 cine maps without extending the scan time. Larger cohorts including patients (e.g., cardiomyopathy patients) are necessary for further investigation.

Acknowledgements

This work was supported by NIH 1R01EB028146.