Combining Parallel Imaging and Model-based Reconstruction for Isotropic 3D T2 mapping with Multi-Echo GRASE
Gian Franco Piredda1,2,3, Tom Hilbert1,2,3, Reto Meuli2, Josef Pfeuffer4, Jean-Philippe Thiran2,3, and Tobias Kober1,2,3

1Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland, 2Department of Radiology, Centre Hospitalier Universitaire Vaudois (CHUV), Lausanne, Switzerland, 3LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, 4Application Development, Siemens Healthcare GmbH, Erlangen, Germany


Classical T2 mapping based on 2D multi-echo spin-echo sequences achieves only limited across-slice resolutions. For clinical use, acquisitions with high isotropic resolution are however desirable, resulting in clinically prohibitive scan times. To this end, we propose a 3D multi-echo gradient and spin echo sequence with CAIPIRINHA and an additional model-based acceleration for T2 estimation. The combination of these techniques allows for whole-brain T2 mapping with 1.6mm-isotropic resolution in 3:26 min. The proposed framework was tested both in phantom and in vivo experiments.


Quantitative T2 mapping of the human brain is typically performed using a 2D multi-echo spin-echo (MESE) sequence1. A mono-exponential decay is voxel-wise fitted to estimate T2 values from images acquired at different echo times (TEs). Interleaved slice sampling is commonly employed to obtain maps within clinically acceptable acquisition times (TA). However, the 2D nature of the MESE sequence and specific absorption rate restrictions of multi-slice acquisitions impede high-resolution isotropic images. Higher and isotropic resolutions could be achieved with 3D MESE acquisitions; however, as 3D prohibits techniques like interleaved slice sampling, TAs become very long.

Here, we investigate brain T2 mapping with an undersampled 3D multi-echo gradient and spin echo (GRASE) sequence2 as an alternative. In the multi-echo GRASE, an echo-planar-imaging (EPI) readout centered on each spin echo is introduced to reduce the TA in comparison to the MESE sequence. Further acceleration combining CAIPIRINHA3 and model-based reconstruction4 is explored, enabling scan times of under four minutes for a 1.6mm-isotropic whole-brain T2 map. The results are compared to reference T2 values estimated from single-echo spin-echo (SE) acquisitions in both phantom and in vivo experiments.


A prototype 3D multi-echo GRASE sequence that supports parallel imaging in both phase-encoding directions was implemented. A separate FLASH scan was acquired as calibration data for the CAIPIRINHA kernels.

All acquisitions were performed at 3T (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany) using a standard 64-channel head/neck coil. Fully sampled and CAIPIRINHA-accelerated (R$$$_y$$$xR$$$_z\!$$$Δk$$$z$$$ = 2x21) multi-echo GRASE data were acquired in a multi-purpose phantom (five compartments with different concentrations of MnCl2$$$\cdot$$$4H2O). After obtaining written informed consent, the CAIPIRINHA-undersampled multi-echo GRASE sequence was scanned in two healthy volunteers (male, 25 and 29 y/o). For both phantom and in vivo experiments, a series of single-slice single-echo SE sequence was acquired with different TEs to compute reference T2 maps by fitting a mono-exponential decay. Relevant sequence parameters are listed in Table 1.

T2 maps were estimated from the multi-echo GRASE data by fitting an extended-phase-graph (EPG) model that assumes only one water compartment5,6. Previous work7 demonstrated that high acceleration of a 2D MESE sequences can be achieved by combining parallel imaging and model-based iterative reconstruction. Following this idea, a combination of CAIPIRINHA and a split algorithm for fast T2 mapping (SAFT)4 was explored. To that end, the CAIPIRINHA-accelerated acquisitions were retrospectively undersampled with a block sampling pattern8 for additional three- and five-fold acceleration. First, CAIPIRINHA kernels were used to fill the missing lines within each sampling block (Figure 1). Subsequently, T2 maps were estimated by minimizing the following cost function:


where $$$A$$$ is the encoding matrix which incorporates coil sensitivities, Fourier transform, and the block-sampling pattern, $$$X$$$ is the current image estimate, $$$Y$$$ the undersampled k-space data, $$$M_0$$$ the initial magnetization, $$$B1$$$ the nominal flip-angle, and $$$\alpha$$$ a regularization parameter. The free variables are estimated by splitting the cost-function and solving it alternatingly4.

T2 values within manually drawn ROIs in the phantom and volunteers were plotted and compared with a two-tailed t-test.


In the phantom experiment, a significant difference (p<0.0025 after Bonferroni’s correction) between reference T2 values and multi-echo GRASE-derived maps with different undersampling schemes was found only in the compartments with the shortest and longest T2 (Figure 2). Notably, the retrospective undersampling of the data and model-based reconstructions did not introduce further variability, except for the compartment with the highest T2.

Representative in vivo T2 maps are shown in Figure 3. Contrary to the phantom experiment, the comparison among manually segmented ROIs in the brain revealed a slight systematic overestimation of the T2 values derived from the multi-echo GRASE data (Figure 4). However, the different undersampling strategies result in similar T2 values.


Experiments conducted both in phantom and in vivo proved the capability of the proposed method to deliver isotropic T2 maps within acceptable clinical acquisition time.

Agreement between T2 values derived from single-echo SE and multi-echo GRASE acquisitions was demonstrated in phantom experiments for the biologically relevant range, even with highly undersampled data.

In vivo data revealed a slight T2 overestimation in multi-echo GRASE-based reconstructions, which seems systematic. This requires further investigation; possible sources include iron deposition, magnetization transfer, or multiple T2 components9. As a next validation step, possible differences between retrospective and real undersampling should also be studied.


This study demonstrated the feasibility of quantitative T2 mapping with highly-undersampled multi-echo GRASE data. An example 60-fold acceleration (3x EPI, 2x2 CAIPIRINHA, 5x model-based) protocol provides whole-brain 1.6mm-isotropic T2 maps in 3:26 minutes.


No acknowledgement found.


1. Carr HY, Purcell EM. Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys Rev. 1954;94(3):630-638.

2. Does MD, Gore JC. Rapid Acquisition Transverse Relaxometric Imaging. J Magn Reson. 2000;147(1):116-120.

3. Breuer FA, Blaimer M, Mueller MF, et al. Controlled aliasing in volumetric parallel imaging (2D CAIPIRINHA). Magn Reson Med. 2006;55(3):549-556.

4. Hilbert T, Thiran J, Meuli R, Krueger G, Kober T. SAFT : Split-Algorithm for Fast T2 Mapping. Proc Intl Soc Mag Reson Med. 2016.

5. Hennig J. Multiecho imaging sequences with low refocusing flip angles. J Magn Reson. 1988;78(3):397-407.

6. Weigel M. Extended phase graphs: Dephasing, RF pulses, and echoes - Pure and simple. J Magn Reson Imaging. 2015;41(2):266-295.

7. Hilbert T, Sumpf TJ, Weiland E, et al. Accelerated T2 mapping combining parallel MRI and model‐based reconstruction: GRAPPATINI. J Magn Reson Imaging. 2018;48:359-368.

8. Sumpf TJ, Uecker M, Boretius S, Frahm J. Model‐based nonlinear inverse reconstruction for T2 mapping using highly undersampled spin‐echo MRI. J Magn Reson Imaging. 2011;34(2):420-428.

9. Malik SJ, Teixeira RPAG, Hajnal J V. Extended phase graph formalism for systems with magnetization transfer and exchange. Magn Reson Med. 2018;80(2):767-779.


Table 1. Parameters of the acquired sequences. The fully sampled multi-echo GRASE was acquired only in the phantom experiment.

Acquisition scheme for a) fully sampled, b) CAIPIRINHA and c) d) CAIPIRINHA combined with block sampling pattern, with green squares indicating acquired samples and orange squares the samples reconstructed with the CAIPIRINHA kernels. The readout direction is orthogonal to the shown planes.

Figure 2. Averaged T2 values in phantom compartments for different sampling schemes. ROIs are shown in the left upper corner. Error bars indicate one standard deviation. Brackets with stars indicate significant difference (p<0.0025 after Bonferroni’s correction for multiple comparisons) between T2 distributions retrieved from reference single-echo SE against multi-echo GRASE-based values.

Figure 3. Axial, sagittal and coronal representative slices of the reconstructed T2 maps in one volunteer for different undersampling schemes. The reference T2 map obtained from a single-slice single-echo SE acquisition is shown in the first column.

Figure 4. Averaged T2 values extracted from manually segmented ROIs in T2 maps of healthy volunteers. The data acquired with the undersampled multi-contrast GRASE sequence show higher T2 values with respect to the reference single-echo SE.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)