Introduction

A challenge for quantitative sodium MRI is the varying transmit and receive B1 fields of the commonly used transmit-receive ²³Na surface coils. Compensating for these varying signal levels requires knowledge of the sensitivity profiles, which are according to reciprocity principle [1] the same for transmit and receive at 33MHz. Transmit B1 can be mapped using the Bloch-Siegert off-resonance pulse approach [2], which is reported to be faster, more robust, and yields higher SNR as to the dual angle approach [3]. The aim of this study is to evaluate Bloch-Siegert off-resonance B1 field correction of sodium images in thigh muscle, heart, kidney, and brain with the use of MRI in healthy human subjects using a 3D FLORET readout trajectory.

Methods

MRI examinations were performed on a 3T MRI scanner (750 MR, GE Healthcare), allowing proton (¹H) MRI as well as sodium (²³Na) imaging. A commercial Helmholtz coil pair (PulseTeq) was used for ²³Na imaging. Scans were performed at the thigh muscle, heart, kidneys, and brain on 5 healthy subjects. The subjects were scanned twice for evaluation of repeatability with minimum two weeks apart. Sodium phantoms were placed in the field-of-view (32 and 80 mmolL^-1, 4% agar) to determine total sodium concentration. The phantoms are placed in the centre of the anterior part of the coil in all scans. Proton anatomical images and B0 maps were acquired.The radio frequency pulses were calibrated (transmit gain and centre frequency) to optimize the sodium imaging acquisition. B1 transmit field map was acquired using a Bloch-Siegert off-resonance approach [2] (+/- 2000 Hz) based on the Fermat Looped, Orthogonally Encoded Trajectories (FLORET) technique [4]. Sodium images were acquired with a 3D-radial-density weighted MR spectroscopic imaging (MRSI) sequence following the trajectory, as described by Nagel et al [5]. Scan parameters of the imaging protocol is listed in Figure 1.
The effect of B1 field correction was evaluated based on anatomical structure, repeatability, and signal homogeneity inside certain organs. Repeatability was estimated from signal-to-noise ratios of high concentration phantoms at initial and follow up scan. Accurate anatomical structure was determined from proton images and homogeneity from standard deviations of histogram probabilities. Multiple paired t-tests were performed on repeatability measures and the standard deviations of histogram probabilities form the original and corrected sodium images. The level of significance was adjusted for multiple comparisons with the Bonferroni-Dunn method.

Results

Visual inspection of the B1 field corrected images indicate an increase of sodium signal. Comparison of anatomy based on the sodium and proton images supports the suggestion towards significant benefits of B1 field correction. This is most pronounced in the brain images (Figure 2, A). However, all anatomies seem to show more truthful structures. This is expected as B1 field correction increases accordingly to distance between coil elements and the region-of-interest (ROI). Blood has a high natural abundance of sodium indicated in signal levels of the two cardiac chambers (Figure 2, B), the bone marrow in the ribs (Figure 2, B & D), and large vessels of the thigh muscle (Figure 2, C).
Sodium phantom signal-to-noise ratio (SNR) were measured in each data set and compared between initial and follow up scan to determine repeatability (Figure 3). Results shows minor changes in SNR when compared to the original images on a group level (Figure 3, C). Furthermore, variation is measured on subject level in both original and corrected images (Figure 3, A+B). Evaluating the variation in multiple paired t-tests with the Bonferroni-Dunn correction showed no significant differences between initial and follow up scan (p < 0.0125).
Quantification of image homogeneity from B1 corrections is determined with a region-of-interest (ROI) at each anatomy (Figure 2). Figure 4 shows the analysis process and difference of applying a histogram (Figure 4, C) as to a cross-sectional line of total sodium concentration (TSC) (Figure 4, F).
The TSC values across matrix row indices shown by the cross-sectional line is challenged when objects with naturally accruing high TSC is crossed. The peak is most likely a blood vessel which has a higher TCS compared to brain tissue makes this approach challenging. Therefore, homogeneity of anatomies was determined from the standard deviations of histogram probabilities (Figure 4, C).
The results indicate a lower variation in the probability of all organs after Bonferroni-Dunn correction (p< 0.0125) this was significant in the brain, heart blood and muscle (Figure 5).

Conclusions

The results illustrate the great potential of the Bloch-Siegert off-resonance based B1 field correction. This is seen in thigh muscle, heart blood, kidney, and brain. The initial benefits were most pronounced in brain. The complete quantitative ²³Na imaging protocol, including B1 field correction is acquired at 3T within 15min. This supports the use of the fast B1 field mapping (2:07min) in the clinical setting.

Acknowledgements

The authors would like to thank radiographer Tau Vendelboe for contributing to scanning of the participating subjects.