Introduction

Electrocardiography (ECG)-gated inversion recovery (IR) acquisition is commonly used in myocardial T₁ mapping¹. Various ECG-gated IR acquisition schemes have been proposed and optimized at different field strengths, mostly based on balanced steady-state free precession (bSSFP) acquisition². Balanced SSFP show advantages in SNR, however, it suffers from banding artifacts due to B₀ inhomogeneities at 3T. Spoiled gradient echo (GRE) is an attractive alternative, especially for volumetric myocardial T₁ mapping at 3T. The robustness of T₁ estimation from spoiled GRE-based ECG-gated IR schemes has not been thoroughly investigated under different schemes and in the presence of B₁ inhomogeneity. In this work, we investigated the effects of B₁ inhomogeneity in the context of T₁ estimation, considering B₁ inhomogeneity in the model for T₁ estimation, and characterized effects of flip angle and heart rate to optimize spoiled GRE-based ECG-gated IR acquisition scheme for 3D cardiac T₁ mapping.

Methods

Simulation Study
Bloch equation simulations were performed to investigate effect of B₁ inhomogeneity on estimated T₁. ECG-gated-IR acquisition schemes of 5-(3)-5-(3) and 8-8 (where N-[M] scheme denotes N number of cardiac cycles for acquisition and M number of cardiac cycles for signal recovery) were investigated for spoiled GRE acquisition at different B₁ scenarios (B₁ = 0.8, 1, 1.2). Monte Carlo simulations were performed with 10,000 iterations and the same noise level across variations. Normalized standard deviation (SD) (i.e., standard deviation of T₁ estimation normalized by standard deviation of noise and square root of acquisition efficiency) was used to assess precision in T₁ estimation for different schemes. ECG-gated IR schemes with variation in numbers of cardiac cycles for acquisition and signal recovery period (N-N, N-[1]-N-[1], N-[2]-N-[2], N-[3]-N-[3] schemes) were investigated for spoiled GRE acquisition. Note the N-N scheme has highest data acquisition efficiency. This is desirable for reducing imaging time, however performance of the scheme on T₁ estimation is unclear. Representative ECG-gated IR schemes of 8-8, 5-(3)-5-(3), and 10-(3)-10-(3) were also investigated for spoiled GRE acquisition with variation in flip angle (from 1-15 in 1° increments). Additionally, 10-(3)-10-(3) scheme was investigated for spoiled GRE acquisition over heart rates from 50 to 120 in 5bpm increments. Simulation details were: heart rate=80bpm, acquisition window=180ms, and TI=100 and 180ms, and flip angle=6°.

Simulation studies utilized a pre-calculated look-up table of Bloch equation simulated signal dynamics to fit the T₁, with and without consideration of B₁ inhomogeneity in the model. The pre-calculated look-up table was generated for a range of T₁ from 1-3000ms in increments of 1ms, and additionally for a range of B₁ from 0-2 in increments of 0.01, when B₁ inhomogeneity was considered.

Phantom and In Vivo Study
Experiments were performed using a 3T MR scanner. For the phantom study, ECG-gated IR schemes of 8-8, 5-(3)-5-(3), 10-(3)-10-(3) were performed at end-diastole with simulated heart rate of 80bpm using spoiled GRE readout. The phantom consisted of 21-vials containing deionized water doped with equally distributed concentrations of gadolinium (Dotarem®) from 0-0.5mM/L. Imaging parameters were: acquisition-window-per-cardiac-cycle=34ms, TI=100 and 180ms, field-of-view (FOV)=360×304mm², matrix-size=192×162, slice-thickness=6mm, TR/TE=3.4/1.9ms and flip-angle=9°. IR with fast spin echo (FSE) readout and MOLLI³ were performed to provide reference and comparison of T₁, respectively.

In vivo experiments were performed on a healthy volunteer using ECG-gated IR scheme 5-(3)-5-(3) at end-diastole with spoiled GRE. Subspace-based data acquisition and image reconstruction methods were used⁴. Imaging parameters were: acquisition-window-per-cardiac-cycle=204ms, TI=100 and 180ms, FOV=360×304×96mm³, matrix-size=192×162×16, TR/TE=3.0/1.5ms, and flip-angle=6°. For comparison, three slices in the apical, mid-cavity, basal regions of the heart were acquired using MOLLI³.

T₁ was estimated for each voxel using cosine similarity between signal dynamics from data and from the pre-calculated look-up table of Bloch equation simulation that includes effects from B₁ inhomogeneity.

Results and Discussion

Simulations from the noiseless case showed bias in T₁ estimation for 5-(3)-5-(3) and 8-8 schemes in the presence of B₁ inhomogeneity (Figs.1A and 2A). This bias was significantly reduced when B₁ inhomogeneity was considered in the model for fitting (Figs.1B and 2B). Simulations with noise showed that normalized SD increased when considering B₁ inhomogeneity in the model for T₁ estimation (Fig.3A). Increasing the period for signal recovery between acquisitions decreased normalized SD, to the level similar to the T₁ estimation case without consideration of B₁ (Fig.3A). 5-(3)-5-(3) and 10-(3)-10-(3) schemes showed overall lower normalized SD compared to the 8-8 scheme, with lowest normalized SD observed for flip angle value around 9° (Fig.3B). For the 10-(3)-10-(3) scheme, similar levels of normalized SD were observed for heart rate variations ranging from 50-120 bpm, for both cases with and without consideration of B₁ inhomogeneity during T₁ estimation (Fig.3C).

Phantom results showed noticeable bias in the estimated T₁ from MOLLI (Fig.4). Overall lower bias and variance was observed in the estimated T₁ from all vials for 5-(3)-5-(3) and 10-(3)-10-(3) schemes using spoiled GRE acquisition compared to MOLLI and the 8-8 scheme (Fig.4). In vivo results showed reconstruction of 3D T₁ maps from 5-(3)-5-(3) schemes using spoiled GRE acquisition with myocardial T₁ comparable to MOLLI (Fig.5).

Conclusion

The accuracy of ECG-gated IR schemes with spoiled GRE readout can be improved by incorporating B₁ into models for T₁ estimation. The precision of ECG-gated IR schemes with spoiled GRE readout can be improved by utilizing an appropriate recovery period.

Acknowledgements

This work was supported in part by the National Institutes of Health (P41EB022544, R01CA165221, R01HL137230, R01HL118261, T32EB013180, and K01EB030045).