Introduction

Sodium is the second most abundant NMR sensitive nucleus in the body. ²³Na MRI of the kidney has been shown to provide the sensitivity to measure the corticomedullary sodium gradient (CMSG), which serves to regulate homeostasis, and has been demonstrated to be reduced in kidney disease and transplanted kidneys^1,2. However, the low sensitivity of ²³Na MRI in the body and the need for additional hardware and specialised RF coils has limited its use. Further benefit of ²³Na MRI comes with the fusion of ²³Na images with ¹H morphological and functional images to produce co-located complementary information, for this the collection of ²³Na and ¹H images within a single scan session without moving the subject is required. Here we compare linear and quadrature ²³Na transmit coils and the use of multi-receive elements for ²³Na of the abdomen using an optimised 3D GRE imaging sequence.

Methods

RF coil design: We collected ²³Na imaging data on a Philips 3T DDAS-MN Ingenia system using two different RF coil designs (Figure 1): 1) a single channel dual loop transmit-receive (Tx/Rx) coil (PulseTeq Ltd., Surrey, UK). The loops are flexible and padded allowing for versatile positioning on a subject. Each loop is placed either side of the imaging region and they act as a Helmholtz pair; 2) a quadrature transmit and 6-channel receive wraparound ²³NA coil (Clinical MR solutions, Brookfield, Wi). The coil comprises of an outside layer containing transmit loops arranged and driven in quadrature, and an inner layer containing two sets of three receive loops. SAR modelling of the coil was performed in REMCOM XFdtd software and RF-induced heating evaluated using a fibre optic thermometer prior to use in-vivo. ¹H images were acquired without moving the subject, using the in-built Q-body coil.
Sequence optimisation: All data acquisition was performed using a 3D GRE sequence, as simulated in Figure 2a. In order to ascertain the optimal combination of parameters a series of 10 x 10 x 30 mm³ scans on phantoms were carried out using pulses with B1 = (5,7.5,10,15,20,30,40)mT at the maximum Ernst angle the average RF power limit would allow. Each scan was carried out twice, one with a single average and the other with a number of averages such that the total acquisition approximately 200 seconds long. The same scans were then carried out in-vivo in a healthy control to validate optimisation.
Data Acquisition: To compare the coils performance a 20-minute ²³Na 3D GRE acquisition was carried out on a 100 mmol/L ²³Na 20L phantom using the following optimum parameters:
Loops: TR: 8ms, TE: 1.75ms, FA: 35, resolution 5x5x20mm³, TA: 20mins, NSA: 406
6-channel: TR: 25ms, TE: 1.97ms, FA: 44, resolution 5x5x20mm³, TA 20mins, NSA: 130
Identical scans were carried out in-vivo to allow direct comparison. In addition, in-vivo a ¹H T1W B-TFE 1.8x1.8x15mm³ scan was acquired for high-resolution segmentation of the kidney to be applied to the ²³Na images, as well as a 12 echo mFFE for renal BOLD.
Data Processing: The six receive channels signals were combined using a sum-of-squares combination³. SNR in the ²³Na image was computed by defining regions of interest (ROI) of similar size and location as the kidneys in the body in the phantom images and comparing it to the signal inside an ROI of background noise.

Results

Results from the parameter optimisation can be seen in Figure 2, it can be see that use of a small TR provides the greatest SNR per unit time for both the loops coil and the 6-channel coil. The loops coil has a higher transmit efficiently meaning the required B1 could be reached with less RF power, allowing use of a shorter TR and higher NSA, while keeping within the average power limit. Figure 3 shows the ²³Na image of the phantom acquired using each coil. The SNR in the ROIs are 6±2 and 19±3 for the loop and 6-channel coil respectively. Figure 4 shows the same acquisition carried out in-vivo using each coil and the associated ¹H image was used to segment the kidneys, the SNR inside the kidneys was 7±3 and 12±4 for the loop and 6-channel coil respectively. Figure 5 demonstrates the FOV of the 6-channel coil and the ability to carry out a T₂* scanning in the same session.

Discussion

The optimum parameters were found for a GRE sequence for ²³Na abdominal imaging. The 6-channel coil provided a higher SNR than the loops coil in both phantom and in-vivo due to the combination of the signal from the individual six receive channels, despite the greater transmit efficiency of the loops coil. Further, the six channels provided a greater FOV and with larger subjects the kidneys would not be within the optimum sensitivity region of the loops coils. The sum-of-square method increases the noise when the channels combined, other methods of channel combination can be used to reduce this effect⁴ which will provide a further increase of SNR. Further investigations into the B1 homogeneity of the two coils is also needed.

Conclusion

Optimised ²³Na imaging of the abdomen has been demonstrated. In future an ultrashort TE sequences⁵ will be implemented to capture the short T₂ ²³Na decay and increase SNR.

Acknowledgements

This work was supported by a Discovery Grant from the Medical Research Council: “²³Na MRI: New frontiers in clinical imaging and diagnostics”