Synopsis

Keywords: Quantitative Imaging, Liver, fat, Dixon

Fat represents a major confounding factor for T1 mapping of the liver. Recently, a number of novel magnetization-prepared multiecho imaging approaches have been proposed to obtain water-only T1 mapping, and they all hold great potential for clinical applications. However, several questions have raised along with this this trend, such as (1) whether concurrent estimation of R2star and the number of echoes is necessary and whether it would affect the quantification of water T1; and (b) whether the number of echoes would affect T1 estimation. This work aimed to investigate these two questions with both numerical simulation and in-vivo experiments.

Introduction

Quantitative measurement of T₁ is highly-desirable in the clinic and could be a valuable biomarker for liver diseases, such as detection of liver cirrhosis and fibrosis^1,2. Unfortunately, T₁ mapping of the liver has not gained clinical acceptance to date due to various reasons. One of them is the influence of fat, which represents a major confounding factor for T₁ mapping of the liver, particularly for fatty livers. In the past few years, a number of novel magnetization-prepared multiecho imaging approaches have been proposed to obtain water-only T₁ mapping^3,4, and they all hold great potential for clinical applications. However, several questions have raised along with this this trend, such as (1) whether concurrent estimation of R₂^* and the number of echoes is necessary and whether it would affect the quantification of water T₁; and (b) whether the number of echoes would affect T₁ estimation. This work aimed to investigate these two questions with both numerical simulation and in-vivo experiments. All the in-vivo experiments uses a recently developed imaging method called Golden-angle RAdial Sparse Parallel T₁ Mapping with Dixon acquisition (GraspT1-Dixon), which enables rapid free-breathing fat/water-separated T₁ mapping in the liver.

Technique and Methods

Signal model
The signal model for a pixel containing water and fat can be written as⁵:
$$s(t)=(We^{-R_{2}^{*}}+F\sum_{p=1}^{P}r_{p}e^{i2\pi{\Delta}f_{p}t}e^{-R_{2}^{*}})e^{i2\pi{\Psi}} (Eq.1)$$
where W and F are the water and fat signal, Ψ is the phase shift caused by static field inhomogeneities. $$${\Delta}f_p$$$ is the p^th fat resonance peak. $$$r_p$$$ is the relative proportion of the p^th fat peak, such that $$$\sum_{p=1}^{P}r_{p}=1$$$.
For inversion recovery (IR)-based T₁ mapping, MR signal can be expressed as:
$$s(t)=A-Be^{\frac{-TI}{T_{1}^{*}}} (Eq.2)$$
Here, $$$A=I_{0}T_{1}^{*}/T_{1}$$$, $$$B=I_{0}(1+T_{1}^{*}/T_{1})$$$, $$$I_0$$$ is the proton density and $$$T_{1}^{*}$$$ is called apparent relaxation rate that is observed when a constant flip angle (FA) is applied repeatedly to the recovering spins. With the estimated three parameters, the true T₁ can then be calculated as $$$T_{1}=T_{1}^{*}(\frac{B}{A}-1)$$$.

Simulation
Simulation using the Bloch equations was conducted to examine the effect of R₂^* decay on the accuracy of T₁ under different fat fractions and signal-to-noise (SNR) levels.
In-Vivo Imaging
In-vivo imaging was performed in 15 volunteers using 2D GraspT1-Dixon, which enables single-shot breath-hold fat/water-separated T₁ mapping. Breath-hold 2D imaging was used to avoid the influence of respiratory motion. Imaging was performed on a Siemens 3T Skyra scanner (Siemens Healthineers, Erlangen Germany) with IRB approval. After one IR pulse, golden-angle rotated six-echo radial spokes were continuously acquired. Relevant imaging parameters included: TE={1.4,3.0,4.6,6.2,7.8,9.4}ms, TR=10.7ms, FOV=360x360mm², matrix size=256x256. 1500 spokes were acquired for each echo in 16s.
Image reconstruction was performed using the GRASP-Pro technique by grouping every 13 consecutive spokes as one image for each TI. Fat/water separation was performed for each TI and a water-specific T1 map was obtained using the three-parameter mode above. For comparison, fat/water separation was performed with and without concurrent R₂^* estimation, so that the resulting water T₁ can be compared. In addition, fat/water separation was also performed using only three echoes without concurrent R₂^* estimation, and the resulting water T₁ maps were compared with that from six echoes.

Results

Fig.1(a) shows that based on the TEs, different R₂^* estimation accuracy were achieved. When the R₂^* is correctly estimated, the T₁ estimation can benefit from R₂^* correction. Simulation was repeated at SNR=30dB for different fat concentrations, results are presented in Fig.1(b). As fat increases, T₁ under-estimation at lower R₂^* gets worse. This is likely due to an inadequate R₂^* value been used in the fat/water separation, the residual fat signal may have lowered the water T₁.

Since the influence from R₂^* decay on T₁ estimation is small, we repeated the simulation at different SNR, the results are presented in Fig.2. The experiments were done with 5% fat fraction. As can be seem, at a higher SNR level, more accurate T₁ can be estimated. At SNR=10dB, R₂^* correction no longer makes a difference.
Bland-Altman analysis was done at the same SNR level with different fat fractions. Experiment was done with R₂^*=75Hz. Results are presented in Fig.3.
Fig.4 presents two examples of in-vivo results. In Fig.4(a), parameter maps from two of the participating subjects, one of which has elevated level of fat in the liver.The composite T₁ maps and the water T₁ maps for the subject with higher liver fat has a large difference. For both subjects, the water specific T₁ maps from different experiments look similar. Results of ROI analysis are shown in Fig.4(b).
Fig.5 presents the in-vivo results for all 15 subjects. Fig.5(a) shows the Bland-Altman plot between T₁ maps from 3-echo and 6-echo sequences, as well as the T₁ maps with and without R₂^* corrections. Fig.5(b) shows the mean T₁ values and the standard deviations across 15 subjects that were analyzed.

Conclusion

The numerical simulation performed in this study has indicated that R₂^* decay does pose an influence on estimation of water-specific T₁. However, this influence is small and can be easily overwhelmed by noise in the images. For in-vivo imaging studies, we were not able to observe the influence of R₂^* decay on water T₁ estimation maps. Six echoes can result in better visual image quality for T₁ maps, but the resulting T₁ values are not significantly different.

Acknowledgements

This work was supported in part by NIH research grants: R01-EB009690, U01-EB029427, R01-EB030549, R21-EB032917 and GE Healthcare.