Quantitative 23Na magnetic resonance imaging in the abdomen at 3 T

Objectives To assess the feasibility of sodium-23 MRI for performing quantitative and non-invasive measurements of total sodium concentration (TSC) and relaxation in a variety of abdominal organs. Materials and methods Proton and sodium imaging of the abdomen was performed in 19 healthy volunteers using a 3D cones sequence and a sodium-tuned 4-rung transmit/receive body coil on a clinical 3 T system. The effects of B1 non-uniformity on TSC measurements were corrected using the double-angle method. The long-component of 23Na T2* relaxation time was measured using a series of variable echo-times. Results The mean and standard deviation of TSC and long-component 23Na T2* values were calculated across the healthy volunteer group in the kidneys, cerebrospinal fluid (CSF), liver, gallbladder, spleen, aorta, and inferior vena cava. Discussion Mean TSC values in the kidneys, liver, and spleen were similar to those reported using 23Na-MRI previously in the literature. Measurements in the CSF and gallbladder were lower, potentially due to the reduced spatial resolution achievable in a clinically acceptable scan time. Mean long-component 23Na T2* values were consistent with previous reports from the kidneys and CSF. Intra-population standard error was larger in smaller, fluid-filled structures due to fluid motion and partial volume effects. Supplementary Information The online version contains supplementary material available at 10.1007/s10334-024-01167-6.


Introduction
Magnetic resonance imaging (MRI) is a powerful tool for routine and exploratory clinical diagnostic imaging of the human body without the need for ionising radiation or invasive biopsies.Whilst routine proton MRI techniques are widely applicable in the imaging of both tissue structure and function owing to the abundance of protons in body fat and water, imaging of other biologically relevant and MR-active X-nuclei such as sodium-23 is also possible.Existing techniques for quantification of sodium content such as flame photometry [1,2], ion-specific electrochemical potentiometric techniques [3,4], or inductively coupled plasma optical emission spectroscopy (ICP-OES) [5] are well-documented in the intracellular and extracellular compartments [6,7], but all require tissue acquisition which is invasive.Non-invasive techniques such as urinary osmolarity have been used to study the corticomedullary gradient (CMG) in the kidneys, but these provide only indirect and non-specific information on kidney function, and are not widely translatable to other organs [8].
In addition to hydrolysis of ATP and the phosphorylation of the Na + /K + pump, endogenous sodium exchange between the intracellular and the extracellular compartments plays an essential role in numerous biological pathways, including regulation of blood volume, pressure and pH, as well as neuronal activation and maintenance of the cell resting potential [9].However, excessive dietary sodium intake has been linked to numerous health risks, with increased blood pressure in particular being a contributing factor in the prevalence of heart disease, chronic kidney disease and stroke among others [10][11][12].Previous histological studies have shown the total sodium concentration (TSC) to be elevated by more than 50% in malignant brain and breast tumours in humans in comparison with healthy tissue, and changes in sodium content are thought to correlate with successful treatment [13,14].In addition, there is a clinical need for the differentiation of tumour types and cellular sodium uptake and retention may provide an opportunity to discriminate more aggressive from less aggressive tumours.
Quantitative analysis of the combined sodium content from both the intracellular and extracellular compartments using MRI has been demonstrated in a number of prior studies [9,[15][16][17][18], and further progress in the field is arising from advancements in dedicated 23 Na-tuned radiofrequency (RF) coil design, pulse sequence development and data analysis [18][19][20][21][22][23][24][25][26][27][28].There is, therefore, an interest in establishing 23 Na-MRI as a quantitative imaging biomarker for cellular integrity and metabolism in disease diagnosis.Recent human imaging studies have included the determination of tumour cellularity in ovarian cancer [29], quantification of sodium in prostate tumours [30], linking of elevated skin sodium content to multiple sclerosis [31] and psoriasis [32], as well as the measurement of tissue sodium accumulation in chronic kidney disease (CKD) [33,34].Ex vivo 23 Na MR spectroscopy of human tissue has also been utilised to observe a reduction in skin sodium content in Type 2 diabetes [35], and determination of the salt-sensitivity of blood pressure (SSBP) using 23 Na MRI may be potentially useful as a non-invasive, quantitative imaging biomarker for hypertension [36,37].
In addition, measurements of the transverse relaxation time (T 2 *) could potentially provide information on tissue oxygenation; measuring the inverse relaxation rate (R 2 * = 1/T 2 *) in proton MR has shown elevation in the presence of paramagnetic deoxyhaemoglobin [38].Quantitative proton T 2 * mapping has been demonstrated as a potential biomarker in the imaging of tumour hypoxia [39,40] and in the prognostic staging of rectal cancer [41].Quantification of 23 Na T 2 * in humans has proven challenging due to additional relaxation effects arising from the quadrupolar nature of the 23 Na nucleus (I = 3/2), where spin relaxation follows a bi-exponential trend with one decay component being substantially longer than the other [49,50].Multiple quantum filtering techniques for separating these decay components have been demonstrated in the human brain, but have yet to find wider application due to the order of magnitude lower signal intensity relative to the single quantum spin-density-weighted imaging [51,52].In addition, the larger and less-efficient coils employed in abdominal imaging necessitate longer pulse widths to achieve desired flip angles, by which point the short T 2 * relaxation component has already decayed to a significant extent, limiting the achievable SNR.To this end, in vivo evaluation of T 2 * estimation in other organs and fluid-filled structures for the first time is of interest to complement previous work in the healthy human kidneys [42,43].
In this study, we report estimates of total 23 Na content (TSC) and long-component transverse relaxation (T 2 *) in eight organs and fluid-filled structures within the abdomen of healthy volunteers (N = 19, 12 M/7F).

Materials and methods
This study was approved by the local ethical review committee (reference number 08-H0311-117).

Data acquisition
Axial 23 Na MR images of the abdomen were acquired using a novel birdcage transmit/receive body coil, possessing a large (~48 cm) field-of-view (FOV) which facilitated uniform excitation over a significant portion of the abdomen in subjects with a range of body habitus, as we have previously described [44].Corresponding axial proton images were acquired at 3 T using a clinical GE MR750 system (GE Healthcare, Waukesha WI).A pair of 50 mL Falcon tubes, each containing an 80 mM agar gel sodium chloride phantom, were suspended from the upper rungs on either side of the 23 Na birdcage coil (Fig. 1) to serve as a calibration standard for signal intensity.In later subjects, two further identical phantoms were placed underneath the patient to the left and right sides of the bed centre, to serve as an additional point of reference.
Healthy volunteers were situated headfirst and supine inside the 23 Na birdcage coil, with the centre of the coil positioned over the kidneys, as confirmed using a fast gradient echo (FGRE) 1 H localiser sequence.Anatomical fat/ Fig. 1 Orientation of healthy volunteers, the 23 Na birdcage T/R body coil and 80 mM agar gel phantoms within the 3 T body MRI scanner water 1 H imaging was achieved using an axial breath-hold 3D T 1 -weighted sequence (LAVA-Flex GRE) with the MR system body coil and 32 locations each (TEs = 1.1/2.2ms, TR = 3.7 ms, flip = 15°, averages = 1, matrix = 256 × 192, phase FOV = 0.7, FOV = 40 cm).Shims were optimized over a large (~16 cm diameter) cuboid region and these shim settings were retained in the following multinuclear scans.
A multinuclear Bloch-Siegert-based calibration scan sequence [45] was prescribed in the centre of the coil to determine the centre frequency of 23 Na and suitable transmit and receive gain values; all values were kept constant for the duration of the scanning session to ensure consistent intensity scaling.A pair of low-resolution images were acquired using a 3D cones trajectory (FOV = 48 cm, TR = 150 ms, TE = 0.705 ms, flip angle = 40/80°, trajectory resolution = 9 × 9 × 9 mm, 197 transients, averages = 3, 125 kHz full receiver bandwidth, total duration = 1:28 min per sequence) to perform B 1 non-uniformity correction using the double-angle method (DAM) [46][47][48].

MATLAB data analysis
All data analysis and region drawing were performed in MATLAB (Mathworks, Natick MA).

B 1 field inhomogeneity correction
For each subject, a map of the relative B 1 value at each voxel was generated from the corresponding intensity ratio of high (80°)-to-low (40°) flip-angle intensity in the two low-resolution DAM series data sets as where I high and I low are the signal intensities at each voxel in the high and low flip-angle data sets, respectively, and α low (1) is the lower of the two flip-angles used (nominally 40° in our protocol).A 3D median filter (kernel size = 5 × 5 × 5 voxels) was applied to maximise the signal-to-noise ratio (SNR) whilst preserving edge information.Example output from the 3D median filtering process in MATLAB is presented in Figure S1 of the Supporting Information (SI).Thresholding was performed, so that the signal intensity ratio was made equal to zero for voxels in the high flip-angle data set, possessing an intensity less than 10% of the maximum observed value.Finally, values were normalised to generate a value of 1000 when the actual and nominal flip angles were equal.
It should be noted that for some of the healthy volunteers, B 1 mapping data were not acquired as part of the imaging protocol.To perform B 1 non-uniformity correction in these subjects, a population average estimate B 1 map was created by taking the median non-zero B 1 value at every voxel across all the remaining subjects (N = 17) and using this as a reference when determining the effective total sodium concentration.
Three-dimensional maps of the B 1 -corrected total tissue sodium concentration (TSC) were calculated from the highresolution 3D cones trajectory series in each of the N = 19 healthy volunteers as 80 mM multiplied by the ratio of the pixel intensity to the average intensity in a region of interest (ROI) drawn inside the calibration phantom.A coronal maximum intensity projection (MIP), shown in Figure S2 of the Supporting Information, was additionally acquired in one subject, but was not used for the purposes of TSC estimation.
However, due to non-linear image distortions which differed between the DAM and high-resolution 23 Na series (shown in Figure S3 of the Supporting Information and discussed in more detail in SI Section 2), a single image transform could not be applied to adequately register both volumes to the 1 H anatomical series.In addition, the tissue contrast of the DAM series was insufficient to reliably register specific regions and structures to the corresponding 1 H anatomical image.Generation of B 1 -corrected TSC maps across the whole abdomen was, therefore, considered unreliable.To this end, TSC calculation was instead performed by undertaking B 1 non-uniformity correction on a per-region basis, drawing a region for each organ or fluid-filled structure in both the high-resolution 23 Na image and the lowresolution DAM B 1 map.Region TSC was then calculated by accounting for the receive and transmit B 1 profiles as follows: where I organ and I ref represent the image intensities in the drawn organ and phantom regions in the high-resolution 23 Na image, whilst B1 ref and B1 organ represent the (2) corresponding normalised intensities in the drawn regions of the B 1 map.C ref is the total sodium concentration of the agar gel phantom used as a reference (80 mM in our study).
In this way, the uncorrected TSC is derived from the first two terms of Eq. ( 2), the receive B 1 correction is applied in the third term, and the transmit B 1 correction applied in the final term.The nominal flip angle α in the high-resolution data set was 70°.
Estimates of TSC from all regions of interest presented in this work have been presented as concentrations in mmol per litre of wet tissue.Absolute quantification of TSC (which would include relaxation corrections) was not performed, although relaxation bias for longitudinal T 1 relaxation and the long component of T 2 * relaxation was assumed to be small based on the repetition time and echo time of RF pulses used in signal acquisition (TR = 100 ms » T 1 and TE = 0.705 ms « T 2 * long ;.More information on relaxation bias estimation can be found in Tables S1 and S2 in Section 3 of the Supporting Information (SI).

Na T 2 * relaxation time constant calculation
Three-dimensional maps of the T 2 * relaxation time constant were calculated from the low-resolution 3D cones trajectory variable TE series in each of the healthy volunteers by taking the negative natural logarithm of the signal intensity at echo times from 2 to 16 ms and performing a linear fit (the first two echo times were not used for analysis, see Figure S4 in the Supporting Information).Thresholding was again performed relative to 10% of the maximum observed signal intensity in the shortest echo time series (TE = 0.705 ms), with intensities below this value set to zero.
For the purposes of this work, estimates of the long T 2 * components only were determined for each organ or structure via log-linear fitting.A comparison of mono-exponential, bi-exponential and log-linear estimation of T 2 * in this body of work is included in Section 4 of the Supporting Information (SI).

Region drawing
Regions of interest were drawn on a single axial slice within MATLAB using the roipoly function to create logical masks over the structure or organ in both data sets, with the mean intensity values in these drawn regions only being used in the TSC and long-component 23 Na T 2 * calculations.Slice selection was performed manually for each organ and volunteer by visually identifying a slice with minimal artifacts and/or partial volume effects from other nearby structures and void spaces.Regions drawn were as follows: phantoms, kidneys, CSF, liver, gallbladder, spleen, aorta, and inferior vena cava (IVC).
As discussed above, the TSC calculation required the drawing of regions within both the high-resolution and lowresolution, DAM data sets to perform B 1 non-uniformity correction.For the long-component T 2 * calculation, regions drawn mirrored those used in the B 1 calculations.Regions were drawn in the image series acquired using the highest flip angle for DAM and the shortest TE for T 2 * to maximise SNR.Regions were drawn of the approximate size, shape, and location as those shown in Fig. 2. In addition, a single large region was drawn outside of the body habitus to perform signal-to-noise ratio (SNR) measurements of uncorrected high-resolution 23 Na signal intensity.

Results
A total of 19 healthy volunteers (12 male, 7 female) underwent imaging for the purposes of this study.

Double-angle method (DAM) B 1 mapping
Example low-flip and high-flip 23 Na images are presented in Fig. 3a, b, respectively, along with a corresponding 1 H anatomical image of summed water and fat shown in Fig. 3c.The resulting B 1 map created from these low-resolution 23 Na images is shown in Fig. 3d, with the population average estimate B 1 map utilised for subjects without DAM acquisition series shown in Fig. 3f).Finally, a high-resolution 23 Na TSC map was generated before applying B 1 correction, as presented in Fig. 3e.All images shown were acquired from a central axial abdominal slice.The population average estimate B 1 map shown in Fig. 3f includes data from all volunteers, where DAM B 1 data were acquired; all other images were acquired in a single healthy male volunteer (age 25-30 years).
Mean SNR of uncorrected 23 Na signal intensity perregion across the healthy volunteer population was highest in the gall bladder (33 ± 15), followed by the CSF (30 ± 4) and kidneys (left: 23 ± 5, right: 26 ± 5).SNR was comparable in the aorta (21 ± 3) and IVC (22 ± 4).The lowest SNR values from uncorrected 23 Na signal intensity images were seen in the spleen (13 ± 3) and liver (11 ± 3).SNR values for individual subjects by region can be found in Table S3 of the Supporting Information.

Total sodium concentration
Intensity maps of acquired 23 Na signal through a healthy male volunteer (age 30-35 years) are shown in Fig. 4. As discussed previously, the mean total sodium content in each organ was not determined directly from the TSC map due to non-linear image distortion between the high-resolution and DAM B 1 23 Na series.B 1 -corrected 23 Na TSC values were calculated independently for each organ or structure in all subjects, and are recorded in Table S3 of the Supporting Information (SI).Median 23 Na TSC values and their associated interquartile ranges are presented in Fig. 5.For all values quoted below, the uncertainty is expressed as the population standard deviation.
Across the 19 healthy volunteers, the 23 Na TSC was highest in the gallbladder with a mean value of 122 ± 42 mM, followed by the CSF, where the estimated mean TSC was calculated to be 94 ± 18 mM.High standard deviation of TSC was also observed in other small, fluid-filled structures, such as the aorta and IVC, where mean TSC values were 78 ± 16 mM and 83 ± 21 mM, respectively.In the kidneys, the coefficient of variation in measured total sodium concentration was somewhat reduced: 79 ± 15 mM and 71 ± 9 mM TSC in the left and right kidneys, respectively.The liver and spleen demonstrated a low sodium content and correspondingly high coefficients of variations, with observed  S4 of the Supporting Information (SI).Mean long-component T 2 * values were highest in the kidneys: 23 ± 3 ms in both cases.Relaxation was observed to be significantly more rapid in the liver with a moderate coefficient of variation (11 ± 3 ms).Whilst exhibiting a higher average long-component T 2 * value than the liver, intra-population variation was similarly modest in the blood: 17 ± 4 ms in the aorta and 16 ± 4 ms in the IVC.Long-component T 2 * relaxation time was observed to be highest in the CSF at 25 ± 5 ms.As with the B 1 -corrected TSC calculations, the most significant intra-population variation in the long-component T 2 * value was typically observed in the gallbladder (16 ± 6 ms), as well as the spleen (10 ± 4 ms).

Double-angle method (DAM) B 1 mapping
The uncorrected 23 Na TSC map through a central axial slice shown in display 3f shows the sharp spatial resolution of the two agar gel phantoms, both kidneys and the CSF and spinal cord, facilitating easy region drawing within MATLAB and quantification of mean 23 Na signal on a per-region basis, as discussed in the following section.
The B 1 maps presented in Fig. 3d, f demonstrate that most organs investigated in this study are within a central region of the abdomen, where B 1 variation in this coil was relatively small: mean fractional B 1 value = 0.97 ± 0.07 in the population average B 1 map.The most significant B 1 variation was localised to structures near the rungs of the birdcage T/R body coil, with maximum fractional intensities as high as double those seen in the central region.Three-plane maps of median fractional B 1 and accompanying standard deviation Fig. 4 23 Na signal intensity maps in axial, sagittal and coronal planes demonstrating sodium content in some major organs and fluid-filled regions of interest in a healthy male volunteer (age 30-35 years) Fig. 5 Total 23 Na concentration box-and-whisker plots illustrating the median (red line), interquartile range (blue box) and maximum extent of non-outlier data points (dashed black lines) for each of the regions of interest interrogated.Whiskers were defined as having a maximum extent of 1.5 times the IQR for each ROI.Note that the values reported for small structures are likely underestimated relative to true concentrations reported in the literature due to partial volume effects across the N = 17 healthy volunteers, where DAM B 1 maps were acquired can be found in Figure S5 of the Supporting Information (SI).
In addition to the large variation in B 1 intensity near the coil rungs which made the correction estimates less reliable, the displacement of the agar gel phantoms used for TSC calibration towards the edge of the FOV could significantly impact TSC calculation due to the prevalence of image distortions inherent in the cones sequence at the FOV edge.Whilst not observed in the healthy volunteer population, to avoid any potential future bias when imaging large participants, this observation led to the inclusion of two additional identical phantoms beneath the patient support, displaced either side of the bed centre.These additional, fixed phantoms could be used as a point of reference in the event of the suspended phantoms yielding poor SNR.

Total sodium concentration
Estimated TSC values in the gall bladder (122 ± 42 mM) and CSF (94 ± 18 mM) were lower than established reports in the literature of ~145 mM sodium concentration in bile [50] and 151 mM in CSF in humans [51] as measured via ion chromatography and those observed in a previous quantification via 23 Na-MRI [52].This discrepancy, as well as high standard deviation of TSC values amongst the healthy volunteer population, was theorised to arise from partial volume effects due to the small region size, since this prior study was conducted using sagittal and coronal plane imaging which presents a significantly larger potential region of interest for CSF than in the axial plane used in this work.Use of a higher spatial resolution would reduce the effects of partial volume errors, but this naturally increases scan time and may impact clinical tolerability.
In contrast, values in the blood vessels were in reasonable agreement with those expected based on a normal extracellular sodium concentration range for healthy adults of 135-145 mM as measured in blood serum [53], coupled with a typical haematocrit fraction or red blood cell volume of ~40-45% in the total blood volume [54].The difference in sodium content within the aorta and IVC was not statistically significant at a 5% level of significance (aorta = 78 ± 16 mM, IVC = 83 ± 21 mM, p = 0.052).
TSC values within the kidneys (left = 79 ± 15 mM, right = 71 ± 9 mM) were around 20 mM lower than those estimated in previous in vivo 23 Na-MRI assessment of the medulla [43] and whole-kidney TSC [55], but were approximately 20 mM higher than those estimated elsewhere using a FLORET B 1 non-uniformity correction method [56].The difference in TSC values observed between the left and right kidneys across our healthy control group was found to be statistically significant at a 5% level of significance (p = 0.029).This unexpected result likely arises from small imperfections in the B 1 correction applied using our low spatial resolution DAM image series.The variation in B 1 is more pronounced near to the kidneys due to proximity of the coil rungs, as visible in Fig. 3f.
The relatively similar standard deviation values for the liver and spleen indicates that they may be primarily driven by the noise floor.The total sodium concentration values in the liver (41 ± 10 mM) were in agreement with previously reported studies in healthy rats [57][58][59], but substantially greater than those observed previously in human studies undertaken using surface coils [60].Flame photometric studies in Dahl rats similarly identified low splenic sodium content relative to other organs [57] (40 ± 9 mM in this work).

23
Na T 2 * relaxation time constant 23 Na T 2 * values from the different abdominal regions generally fell within the range of previously reported long 23 Na T 2 * values [9,61].
The low observed intra-population standard deviation of long-component T 2 * in the kidneys was expected for large organs with high relative 23 Na signal intensity, owing to reduced variability resulting from partial volume effects.Measurements of 23 Na T 2 * in the kidneys were in agreement with previous reports of 21 ± 1 ms at 3 T [43] and 7 T [42,62].Unlike the renal TSC values, the difference in T 2 * relaxation times observed between the left and right kidneys was not affected by B 1 variation, and was not statistically significant at a 5% level of significance (left kidney = 23.2 ± 2.5 ms, right kidney = 22.7 ± 3.1 ms, p = 0.644).
The relatively short T 2 * values observed in the liver (10.7 ± 2.7 ms) are in agreement with previous 1 H-MRI studies which identified a significantly increased component of short-T 2 * tissue in the liver relative to long-T 2 * counterparts [63][64][65], but are lower than those reported in rat liver [66].This observation is thought to arise predominantly from the high content of paramagnetic iron within the liver, which has the effect of increasing 23 Na relaxation rates [67,68].
Observations of 23 Na T 2 * in the blood were contrary to the expectations that both the rate of relaxation and the intra-population variation would be higher due to partial volume effects and the influence of fluid flow given the spatial resolution used for acquisition [69][70][71].Oxygenated (aorta) and deoxygenated (IVC) blood relaxation rates in vivo were observed to be statistically different at a 5% level of significance (aorta = 17.3 ± 4.2 ms, IVC = 15.7 ± 4.0 ms, p = 0.013).This observation was expected considering the different magnetic properties of oxyhaemoglobin (diamagnetic) and deoxyhaemoglobin (weakly paramagnetic), as demonstrated by the higher signal intensity of oxygenated blood in T 2 *-weighted BOLD fMRI [72,73].
Fluid motion effects on T 2 * relaxation in the CSF were not considered to be significant due to the slower rate of CSF motion relative to blood [74] and compared to the spatial imaging resolution employed here.However, the low spatial resolution may explain why the calculated values (24.6 ± 4.6 ms in this work) were significantly below those reported in studies of the brain at 7 T [75].23 Na T 2 * estimates in the gall bladder were limited by the small region size, whereas quantification in the spleen suffered from an inherently low relative 23 Na signal intensity.Observation of the shortest long-component T 2 * values in the spleen is consistent with reports in a murine model comparing relaxation properties with that of the heart and liver [76], although in that study, a statistically significant difference between spleen and liver relaxation rates was noted, which is not observed here (gall bladder = 15.8 ± 6.0 ms, spleen = 10.3 ± 4.2 ms, p = 0.188).Human studies of 1 H transverse relaxation rates also highlighted faster relaxation in the liver than in the spleen [77] which has not been observed for 23 Na.
A comparison of results observed in this work with those published elsewhere in the literature is summarised in Table 1 for reference.

Conclusions
We have demonstrated the potential for acquisition of 23 Na signal intensity across a large proportion of the abdomen using 23 Na-MRI and a large-FOV sodium-tuned T/R body coil with good signal-to-noise ratio in a clinically acceptable timespan.Subsequent estimation of DAM B 1 -corrected total sodium concentration and the long-component of the transverse 23 Na T 2 * relaxation time in a variety of abdominal organs and fluid-filled structures has been demonstrated.This work will serve as a baseline to evaluate pathological changes in conditions such as hypertension and for tumour characterisation.It is hoped that the development of 23 Na-MRI as a biomarker for tissue sodium content and dynamics may in the future inform clinical management by reducing the need for invasive biopsy techniques.
Two identical 80 mM agar gel sodium chloride phantoms were used to estimate TSC in this work to serve as a reliability test, since differences in phantom location can result in variation in local B 1 , and may also experience variation in RF artifacts.In addition, phantoms may dry out over time, increasing their effective sodium concentration.An alternative method for estimation of 23 Na TSC which has been demonstrated in studies in the leg [78] and intervertebral discs [52] and could be applied to future work makes use of multiple agar gel phantoms of different concentrations and performing a linear fit through the 23 Na signal intensities of each phantom and the organ or structure of interest.Presumably, as in these studies, the phantom arrangement should also occupy a fixed position within the coil, such as underneath the patient, to make phantom B 1 calibration measurements more consistent, and phantoms could also be larger in size to facilitate easier region drawing.This may be particularly important in larger patients, as the suspendedphantom arrangement utilised in this work may be less reliable when phantoms are displaced close to the rungs of the coil or the edge of the field-of-view, but at the same time should not impede patient comfort.Estimation of 23 Na T 2 * measurements in fluid-filled regions might be made more reliable using cardiac-triggered acquisition to suppress the effects of fluid motion, but this would have the consequence of significantly increasing scan durations.
Analysis of anatomical variation in TSC and correlations with other MR-based metrics could be facilitated by automated registration of anatomical 1 H, low-resolution DAM B 1 and uncorrected high-resolution TSC images.This may be made easier with the use of machine-learning or deeplearning registration techniques, to overcome the difficulties presented by image rotations and distortions prevalent with conical trajectories possessing long readout lengths.
if changes were made.The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.

Ferdia A .
Gallagher and Mary A. McLean have contributed equally to this work.

Fig. 2 a
Fig. 2 a Example images acquired from high-resolution 3D conical trajectory axial slices in a female healthy volunteer (age 25-30 years), illustrating the region-drawing process in MATLAB: a both kidneys and both phantoms; b CSF; c gallbladder; d spleen, liver, aorta, and inferior vena cava (IVC)

Fig. 3 23 Na T 2 *
Fig. 3 Example sodium images.a Low-resolution 23 Na signal intensity image at a 40° flip angle; b low-resolution 23 Na signal intensity image at an 80° flip angle; c fused fat/water anatomical 1 H image (acquisition FOV = 40 cm, scaled to match other displays); d relative B 1 intensity map calculated using the double-angle method; e high-resolution uncorrected 23 Na signal intensity map; f population average estimate B 1 intensity map.All images were acquired from

Fig. 6 aFig. 7 23
Fig. 6 a Anatomical 1 H T 1 -weighted image through a central axial slice centred on the kidneys for a healthy female volunteer (age 35-40 years).A corresponding axial low-resolution 23 Na image from