Introduction

T₂ relaxation time is one of the main tissue properties in use today in MRI diagnostics. It is of special interest in multiple sclerosis and other demyelinating diseases, and of further appeal by the improved sensitivity at ultra-high field MRI (≥7T). There is a bulk of studies exploring spin-echo and multi-echo based methods to generate T₂ weighted imaging and T₂ mapping^1,2. However, the challenges at 7T MRI include an increase in SAR – resulting in long scans – as well as RF (B₁) inhomogeneity that complicates the interpretation of T₂ weighted images. Steady-state short TR methods based on GRE are also well studied and are used to provide T₂ maps, however, they are more used for combined T₁ and T₂ quantifications^3,4,5. A recent study at 3T demonstrated that phase images from two 3D-GRE scans with specific RF pulse phase increments can be used to produce T2 maps⁶. The phase of this signal, however, also depends on the flip angle. In this study, we propose to acquire four scans (with two flip angles), from which we can estimate both the T₂ and the B₁ maps. We denote this method as Steady-state T₂ And Rf Estimation (STARE).

Methods

The method is based on the well known RF spoiling used with spoiled GRE (SGRE), which can be acquired with short TR (in this work TR of 10 msec was used). However, instead of the standard phase increments (φ_inc) of 117 as used in RF spoiling, small phase increments (for example, φinc) provide notable T₂ contrast in the phase images. Detailed Bloch simulations have shown that the phase of the acquired signal (θ_sig ) includes a global phase, dependent on B₀ and B₁ contributions, as well as a phase mainly dependent on T₂ and flip angle (α): θ_sig = θ(T₂, α) + θ0 . (A small T₁ dependency also exists, however, it is negligible in the relevant range). If two scans are acquired with φ_inc=2 and φ_inc=-2, they provide the same θ₀ , while θ(T₂, α)_φinc=2=- θ(T₂, α)_φinc =-2. If we acquire four scans, two pairs of φ_inc=2 and φ_inc=-2 with two flip angles (for example, α and 2α), we can eliminate θ₀ from the pairs φ_inc=2 and φ_inc=-2 and using Bloch simulations estimate the T₂ and B₁ distribution – based on a dictionary (by closest distance). Figure 1 shows Bloch simulation of the θ(T₂, α) at φ_inc=2. One can also acquire just three scans – 1) α , φ_inc=2; 2) α, φ_inc=-2; and 3) 2α, φ_inc=2 . Then, from 1)+2) calculate θ₀ and use data from 3) to estimate T₂ and B₁ maps, thus further accelerating the acquisition. In this feasibility study, we used four scans. 7T imaging was performed on a 7T MAGNETOM Terra (Siemens Healthcare, Erlangen, Germany) and one phantom dataset was acquired on a 3T MAGNETOM Prisma (Siemens Healthcare, Erlangen, Germany). Three tubes with different agar concentrations (1.5%, 2 .5% and 3.5%) were used to compare the T₂ estimation to the gold standard, spin echo for a set of TE values. A cylindrical phantom with 2% uniform agarose concentration and 0.5% NaCl was examined in both 7T and 3T to inspect the T₂ and B₁ distribution. Finally, human scanning with 1.5 in-plane resolution was acquired and T₂ weighted images and T₂ maps were generated. In-vivo 3D-GRE scan parameters were Sagittal plane, FOV 216x220x128 mm³, resolution 1.5x1.5x2 mm³, TR/TE 10/4.3 ms, α=15° and 26°, scan duration 1:30 min, four scans 6 minutes.

Results

Fig. 2 shows results of T₂ quantification on the three examined tubes with STARE. The scans included α=15° and α=29°, φ_inc=2 and φ_inc=-2. The figure shows the raw data magnitude images, θ distribution (after θ₀ removal), estimated T₂ map, and a comparison to the spin-echo scan, showing a very good fit. Fig. 3 shows quantitative results for the central slice of the cylindrical phantom, comparing 3T and 7T acquisition. A uniform T₂ was reconstructed in both cases, with a pronounced B₁ distribution at 7T, as expected. Fig. 4 shows human volunteer results – where one can see low contrast magnitude images, but clear T₂ weighting in both the estimated θ and in the imaginary signal . The low B₁ values in the cerebellum, as expected, lead to lower θ values compared to other regions with the same T₂. Nevertheless, solving for B₁, in addition to T₂, allows for an overall reliable T₂ estimate.. Fig. 5 shows a set of T₂ weighted images based on the T₂ maps found in Fig.4 for TE=30 ms.

Conclusions

STARE offers a new capability to acquire fast 3D dataset for T₂ mapping, based on a simple 3D-GRE acquisition. This method is based on the phase images and has the advantage of fast acquisitions of a whole brain and the ability to separate T₂ and RF contributions. In this feasibility study, acceleration of the acquisition was not utilized, however, both multi-channel acceleration methods (SENSE/GRAPPA) and Compressed Sensing methods can be easily incorporated in this sequence. Further study is required to analyze its sensitivities and accuracy and optimize φ_inc and α. Note, the method is currently more limited at high T₂ values and very low flip angles.

Acknowledgements

No acknowledgement found.

References

References [1] Juras V et.al. Eur Radio. 2016;26, [2] Brittain JH et.al. MRM 1995; [3] Deoni SCL, MRM 2003; 49; [4] Welsch GH et.al. MRM 2009;62; [5] Leroi L. et.al. MRM 2020; [6] Wang X. et.al. MRM 2019; [7] Pruessmann KP. Et.al. MRM 1999 42, [8] Griswold MA et.al MRM 2002 47, [9] Lustig M et.al MRM 2007 58;