Ultrashort TE 3D spectroscopic imaging for high SNR imaging and bi-exponential signal decay characterization of sodium
Jetse S. van Gorp1, Paul W. de Bruin2, and Peter R. Seevinck1

1Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands, 2Radiology, Leiden University Medical Center, Leiden, Netherlands

Synopsis

Sodium relaxation behavior has been related to structural and cellular integrity, which is of interest for early disease detection. However, the short T2* values, bi-exponential relaxation behavior and low sensitivity makes accurate signal characterization challenging. In this work a 3D-UTE-FID-SI sequence was developed to measure the sodium decay curve with a 32kHz temporal resolution and sub-ms TE to characterize the bi-exponential signal decay characteristics of sodium in vitro and in vivo.


Purpose

To develop a robust quantitative sodium (23Na) MR imaging strategy for sodium imaging with high sensitivity and simultaneous characterization of the bi-exponential 23Na signal decay.


Introduction

Sodium relaxation behavior has been related to structural and cellular integrity, which is of interest for early disease detection1,2. In addition to total sodium concentration measurements it is possible to also detect bound and free sodium pools3, which is expected to improve the specificity of disease detection2. The bound and free sodium pool can be distinguished by their bi-exponential and mono-exponential decay properties, respectively. The bi-exponential decay consists of a long (10-65 ms) and short T2 component (0.5-3.0 ms), which require sequences with ultra-short TE’s for accurate signal characterization. In this work we developed a 3D-UTE-SI approach4 to exploit the favorable properties of spectroscopic imaging for bi-exponential signal decay characterization and efficient sodium imaging.

Methods

Sequences and hardware: Single quantum (SQ) signal decay curves were sampled with high temporal resolution (32kHz) and a sub-ms TE (0.6 ms) using 3D spectroscopic imaging (SI) with variable echo-time sampling (UTE-SI) (fig 1) on a 7T scanner equipped with an in-house built 23Na birdcage coil with 4 1H-strip lines5.

Simulations: 2D-FID-SI simulations were performed using JEMRIS6 with TE/TR=0.6/100ms for T2 components of 0.5ms and 11.4ms with 20ms T1, corresponding to relaxation times found in cartilage7. The time domain signal was used to simulate the effect of T2* decay on different k-space trajectories (UTE-SI, Cartesian and spiral (4 or single shot)) (fig 2) by assigning different TE values to each k-space point corresponding to the k-space sampling order.

Experiments: First a 3D-SI phantom measurement with TE/TR=1.25/133ms was performed to confirm bi-exponential signal decay in a 4% agarose phantom with sodium concentrations ranging between 50-250mM and a tube containing 100mM sodium in H2O (fig 3). Other imaging parameters included: 43 mm3 voxels, 1283 mm3 FOV, 900 block pulse, 4096 samples, 36 min acquisition time. T2* mapping was performed using a constrained least-squares bi-exponential fitting routine in Matlab with cutoff values of the decay curves at SNR=3. The decay curves were resampled to a bandwidth of choice to increase the SNR and robustness of fitting. Additional phantom data with 2x2x4mm3 resolution and 1283 mm3 FOV were acquired in 2x25min for the 3D-SI (TE=1.98ms) and 3D-UTE-SI (TE=0.6ms) data to demonstrate the differences between both sequences (fig 4). Additional imaging parameters included: TR= 22ms, FA 570, 512 samples. Finally, 3D-UTE-SI in vivo data was acquired in the knee of a healthy volunteer for the purpose of T2*short and T2*long determination and high SNR 23Na imaging (fig 5). The data was acquired in 25min with TE/TR=0.6/22ms, 3x3x5mm voxels and FA 700. All data was acquired using a 3D spherical shutter to reduce the acquisition time a factor 2 and 32kHz data sampling


Results

The simulation results (fig 2) demonstrated increased blurring due to T2* effects during k-space traversal. These blurring effects become larger for longer k-space trajectories (fig 2c-e) and shorter T2* values (fig 2 bottom row). Ideally a conventional SI sequence is used to obtain the lowest possible blurring (fig 2a), however the shortest achievable TE is limited by the gradient switching times. By using the shortest possible encoding time for separate layers in k-space with the UTE-SI sequence (fig 1b,2b) the blurring can be minimized without significant loss of signal due to T2* decay, in contrast to Cartesian (fig 2c) and spiral trajectories (fig 2d-e).

The 3D-(UTE-)SI phantom experiments provided high SNR 23Na images (fig 3a, 4) and densely sampled decay curves (fig 3b). The bi-exponential behavior of the sodium signal was clearly demonstrated by non-linear behavior in the log magnitude plot (fig 3c) and quantitatively confirmed in the fitting procedure (fig 3d). By voxel-wise fitting of the data it was possible to reconstruct T2,short* (fig 3e) and T2,long* (fig 3f) maps and obtain the amplitude ratios (fig 3g-h), which were close to the theoretical approximation of 0.6/0.43.

Shortening of the TE to 0.6ms (fig 4a) from 1.98ms (fig 4b) with the UTE-SI sequence enabled bi-exponential T2* decay detection in vivo (fig 5b-c) where very short T2* values of ~0.3-1ms were detected. Analogous to the phantom experiments T2* maps and amplitude maps were derived (fig 5d-g), illustrating bound sodium fractions in cartilage.


Discussion

The proposed 3D-UTE-SI method can be used to study the bi-exponential relaxation behavior of sodium with a sub-ms TE and produce high SNR images for quantitative sodium imaging. This sequence can be used to test underlying assumptions on the sodium relaxation behavior in biological tissues with a high temporal resolution.

Acknowledgements

References

1. Ling, Magn. Reson. Med. 56:1151-1155 (2006), 2. Qian, Magn. Reson. Med. 74:162-174 (2014), 3. Bull J. Magn. Reson. 8:344-353 (1972), 4. Robson, Magn. Reson. Med. 53:267:274 (2005), 5. De Bruin, NMR in Biomed. DOI:10.1002/nbm.3368 (2015), 6. Stöcker, Magn. Reson. Med. 64:186-193 (2010), 7. Madelin, NMR in Biomed. 25:530-537 (2010)

Figures

Figure 1: 3D-FID-SI (a) and 3D-UTE-FID-SI (b) k-space sampling coverage as a function of TE. The main difference between both sequences is an equal TE for all k-space points for conventional SI (a) compared to the varying TE over k-space for the UTE implementation (b).



Figure 2: 2D simulations of a sphere with 11.4 ms T2 (top rows) and 0.5 ms T2 (bottom rows) were performed. A 2D-FID-SI sequence with 16kHz sampling rate (a) was used to reconstruct UTE-FID-SI sampling (b), Cartesian sampling (c) and spiral sampling with 4 shots (d) or a single shot (e).



Figure 3: 3D-FID-SI decay curves of a phantom with varying concentrations (a) demonstrates the bi-exponential relaxation behavior of sodium in agarose gels (b-d) and the mono-exponential decay in H2O (center tube). All individual decay signals were fitted with a bi-exponential decay model to obtain T2*,short (e), T2*, long (f) and M0,short (g) and M0,long (h) maps.



Figure 4: 3D-UTE-FID-SI (a) compared to 3D-FID-SI (b) sodium imaging and their difference images (c) for the minimum achievable TE.



Figure 5: An in vivo 3D-UTE-FID-SI dataset was used to obtain high SNR sodium images (sampling time of 11ms) (a), decay curves (b-c), T2*short and T2*long maps (d-e, note the different scale) and the relative amplitude contributions of the long and short component (f-g) after applying a spatial Gaussian filter.





Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2364