Introduction

Quantitative T₂ mapping offers useful tissue characterization for a variety of clinical and basic research, including estimation of iron overload, stroke severity¹, cartilage and demyelinating diseases among which is multiple sclerosis^2-4. Implementing quantitative T₂ mapping at ultra-high field (≥7T) MRI can offer increased sensitivity, however, still poses a great challenge. Common T₂ techniques at 7T suffer from high SAR and from high RF field (B₁) inhomogeneity, resulting in prolonged scan duration and reduced estimation quality. Recent work⁵ at 3T has demonstrated that T₂ maps can be estimated from phase images of two (modified) 3D spoiled GRE scans with short TR. As a follow-up, we recently presented an estimation method of both T₂ and B₁– denoted as Steady-state T₂ And RF Estimation (STARE)⁶. This method was based on four scans with two flip angle values and overcomes the 7T RF field inhomogeneity, offering a fast 3D T₂ and B₁ mapping for 7T brain imaging. In the current study, we propose to shorten the total scan duration by introducing a dual-steady-state with interleaved flip angles, which we denote as TWISTARE. In this approach, we can extract the T₂ and RF maps from only two scans.

Methods

The method is based on the well-known RF spoiling used with spoiled GRE which can be acquired with short TR (TR of 15 msec in this work). However, instead of a standard phase increment (φ_inc), that dephases all transverse magnetization, here a small phase increment (e.g., φ_inc=2°) introduces coherent transverse magnetization and provides a notable T₂ contrast in the phase images. Bloch simulations show that the signal’s phase ($$$\angle$$$S) includes B₀-dependent phase (θ₀) along with a dependence on T_2, flip angle (α), and φ_inc: $$$\angle$$$S=θ(T2,α,φ_inc)+θ_{0 .} α here is the actual local flip angle, due to the requested α_scan and the local B₁ distribution. (A small T₁ dependency also exists, however, it is negligible in the relevant range). In this study, a dual-steady-state is implemented by interleaving two flip angle values– odd excitations with α_scan, even excitations with AR∙α_scan (AR-amplitude ratio)– producing odd and even phase maps- $$$\angle$$$S^odd,$$$\angle$$$S^even– see the simplified pulse sequence in Fig.1. Such a dual-steady-state is useful for removing θ₀ while keeping T₂ and α dependency: ∆θ(T₂,α,AR,φ_inc)=$$$\angle$$$S^odd-$$$\angle$$$S^even=θ^odd(T₂,α,AR,φ_inc)-θ^even(T₂,α,AR,φ_inc). After repeating the scan with another set of (φ_inc,α_scan,AR) two phase-difference maps -∆θ₁,∆θ₂ - are available and can be exploited for T₂ and α estimation. The two sets of scans parameters –φ_inc1,α_scan1,AR₁ for scan 1 and φ_inc2,α_scan2,AR₂ for scan 2 – define the precision and accuracy of the method. We performed a set of Bloch simulations and an initial SNR analysis to optimize the sensitivity for relevant T₂ and α distribution in 7T brain imaging. The current chosen set of parameters was α_scan=21°,AR=1.5,φ_inc=1.5 for scan 1, and α_scan=12°,AR=3,φ_inc=7.5 for scan 2. Using Bloch simulations we introduce a new 2D space (∆θ₁,∆θ₂) in which for each measured (∆θ₁,∆θ₂) we can estimate T₂(∆θ₁,∆θ₂) and α(∆θ₁,∆θ₂) by interpolation. Fig.1 shows the flowchart of the method, while Fig.2 shows representative Bloch simulation results for a single-steady-state and a dual-steady-state phase dependence.
Phantom and human imaging was performed on a 7T MAGNETOM Terra (Siemens Healthcare, Erlangen, Germany). The quality of the T₂ and α estimation was assessed using a 3D head-shaped phantom⁷ with a similar-to-human RF field inhomogeneity and a uniform T₂ value. Phantom scan parameters were: Sagittal plane, FOV 220x225x256 mm³, resolution 1.3x1.3x2 mm³, TR/TE 15/1.67 ms, single scan duration 5:47 min, total scan duration 11:34 minutes. In-vivo scan parameters were Sagittal plane, FOV 220x225x216 mm³, resolution 1.25x1.25x1.5 mm³, TR/TE 15/1.67 ms, elliptical sampling, single scan duration 5:06 min, total scan duration 10:12 minutes. We used the BART⁸ software to reconstruct the final images for in-vivo dataset.

Results

Fig.3 shows the simulated equi-T₂ and equi-α lines in the (∆θ₁,∆θ₂) 2D space for the selected TWISTARE set of parameters. Excluding extremely low α values, the equi-T₂ and equi-α lines are nearly orthogonal, which allows robust estimation for a given set of (∆θ₁,∆θ₂)_. Fig.4 shows the phantom imaging results– with good fit to the vendor RF field map. The T₂ map shows uniform distribution inside the “brain” compartment, as expected. Since the T₂ distribution in this phantom is uniform, the precision of the method was evaluated by calculating the STD of the T₂ estimate over a large centered box (see Fig.4c), resulting in a STD of 2.17 msec. Fig.5 shows volunteer imaging results of both T₂ and α maps. The α map is similar to the vendor RF map distribution (with some deviations near the ventricles).

Conclusions

The TWISTARE method offers a fast 3D T₂ and α mapping in 7T brain imaging. In this study, acceleration was only partially utilized using elliptical acquisition available in 3D-GRE and hence, additional acceleration can be easily incorporated to further shorten scan duration. The method fits the expected results and shows good precision in phantom imaging. Detailed examination of practical precision, accuracy and replicability with optimized set of parameters is required. Further study is required to compare the estimated T₂ maps to the gold standard spin-echo method.

Acknowledgements

No acknowledgement found.

References

[1] Siemonsen S, Mouridsen K, Holst B, et al. Stroke. 2009; [2] Bruschi, N., Boffa, G. & Inglese, M. Eur Radio. 2020; [3] Henry, T. R. et al. Radiology 2011; [4] Massire, A. et al. Magn. Reson. Med. 2015; [5] Wang X. et.al. MRM 2019; [6] Schmidt R. et.al. Proc. Intl. Soc. Mag. Reson. Med. 29 (2021), 0551; [7] Jona et.al. NMR BioMed 2020; [8] https://mrirecon.github.io/bart