Quantitative 23 Na-MRI of the intervertebral disk at 3 T.

Monitoring the tissue sodium content (TSC) in the intervertebral disk geometry noninvasively by MRI is a sensitive measure to estimate changes in the proteoglycan content of the intervertebral disk, which is a biomarker of degenerative disk disease (DDD) and of lumbar back pain (LBP). However, application of quantitative sodium concentration measurements in 23 Na-MRI is highly challenging due to the lower in vivo concentrations and smaller gyromagnetic ratio, ultimately yielding much smaller signal relative to 1 H-MRI. Moreover, imaging the intervertebral disk geometry imposes higher demands, mainly because the necessary RF volume coils produce highly inhomogeneous transmit field patterns. For an accurate absolute quantification of TSC in the intervertebral disks, the B1 field variations have to be mitigated. In this study, we report for the first time quantitative sodium concentration in the intervertebral disks at clinical field strengths (3 T) by deploying 23 Na-MRI in healthy human subjects. The sodium B1 maps were calculated by using the double-angle method and a double-tuned (1 H/23 Na) transceive chest coil, and the individual effects of the variation in the B1 field patterns in tissue sodium quantification were calculated. Phantom measurements were conducted to evaluate the quality of the Na-weighted images and B1 mapping. Depending on the disk position, the sodium concentration was calculated as 161.6 mmol/L-347 mmol/L, and the mean sodium concentration of the intervertebral disks varies between 254.6 ± 54 mmol/L and 290.1 ± 39 mmol/L. A smoothing effect of the B1 correction on the sodium concentration maps was observed, such that the standard deviation of the mean sodium concentration was significantly reduced with B1 mitigation. The results of this work provide an improved integration of quantitative 23 Na-MRI into clinical studies in intervertebral disks such as degenerative disk disease and establish alternative scoring schemes to existing morphological scoring such as the Pfirrmann score.

gradient echo

proteoglycan

region of interest

tissue sodium content

INTRODUCTION

Back pain is a common medical symptom with massive socioeconomic implications due to effects on the patients' well‐being, the inability to work, and the consequential high healthcare costs. Conventional 1H‐MRI with morphological sequences is a routinely used imaging technique and provides a tool to rule out common causes such as disk extrusion, degenerative alterations of the facet joints or nerve root compression. However, the symptoms are often unrelated to degenerative findings. ,

Noninvasive quantification of the local sodium content by MRI is a sensitive measure of tissue integrity and a valuable method to monitor the tissue viability and ion homeostasis in clinical research. Both the technological advancements on the hardware side of MRI and the emergence of efficient data acquisition techniques (such as ultra‐short echo‐time—T E—sequences) have enabled the utilization of 23Na‐MRI as a potential tool to directly infer functional and structural information on the tissue, otherwise not achievable by conventional 1H‐MRI. , ,

Previous studies have shown that the sodium content in the disks correlates with the proteoglycan (PG) content, which in turn seems to be a biomarker of degenerative disk disease (DDD) and of lumbar back pain (LBP). , Histologically, the intervertebral disk consists of an outer annulus fibrosus and an inner nucleus pulposus. Biochemically, these components are largely composed of collagen and PG. The negative fixed charge density attracts sodium ions, resulting in the concentration of sodium within the disk being directly proportional to the fixed charge density. The pathologic and degenerative reduction of PG content and attraction of positively charged sodium ions causes a lower oncotic pressure and ultimately collagen degeneration. Therefore, 23Na‐MRI is a sensitive method to estimate the changes in the PG content of the intervertebral disk, similar to the articular cartilage. ,

Although 23Na‐MRI offers valuable supplementary information to 1H‐MRI, the practical implementation of tissue sodium quantification poses high demands on the measurement accuracy and precision. First, the signal‐to‐noise ratio of 23Na‐MRI is significantly lower than that of 1H‐MRI. This is mainly because the sodium concentration in the tissue is much lower (45 mM to 350 mM) relative to the water concentration and the gyromagnetic ratio is smaller ( ), ultimately yielding an 11 000 times smaller 23Na signal relative to 1H‐MRI, under equivalent detection conditions for the two nuclei. Second, as a spin‐3/2 nucleus, 23Na signal poses rapid biexponential transverse relaxation and may also exhibit energy eigenstate shifting (or residual quadrupole splitting) as a result of the local electric field gradients due to the long‐lived spatial orientation of the nuclear electric quadrupole moment. , Third, blurring results from the point spread function and from physiological motion. Fourth, strong partial volume effects because of the large voxel sizes lead to further inaccuracies. Finally, the signal modulation due the inhomogeneity in the field appears as a significant challenge for the accurate quantification. , The quantification of the tissue sodium content (TSC) in units of millimoles per liter potentially allows the inter‐ and intra‐individual comparability necessary for patient stratification and for therapy monitoring. However, quantitative analysis of sodium concentration from 23Na‐MRI requires accurate control of factors modulating the sodium signal.

Primarily, the accuracy and precision of the quantitative tissue sodium measurement are highly dependent on the spatial and temporal modulation of all the magnetic fields involved in the experimental realization. Significant error can be introduced due to inhomogeneity in TSC measurement. , A common approach to overcome this issue in 1H‐MRI is to use a relatively homogeneous transmit coil (such as a quadrature birdcage) by assuming that variations in flip angle are small in the imaging volume. However, if body‐volume coils or other highly inhomogeneous transmit coils must be used, this is not applicable and inhomogeneity has to be included in the quantification model. The specific geometry of the intervertebral disks and necessary antenna specs to image this geometry push the accuracy demand in quantification. Therefore, particularly for the intervertebral disks, field variations have to be mitigated by adequate correction methods to perform a reliable tissue sodium quantification.

The goal of this work is to provide quantitative TSC measurements using 23Na‐MRI in a clinical setting by taking into account the effects of spatial variations in the field along and across the intervertebral disk anatomy. To this end, 23Na‐MRI was performed to image the intervertebral disks at 3 T in healthy human subjects. The sodium maps were acquired with a double‐tuned (1H/23Na) transceive chest coil and their individual effects in tissue sodium quantification were examined.

EXPERIMENTAL

MRI hardware

All measurements were made on a 3 T whole‐body MRI scanner (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany). A dual‐tunable (1H/23Na) chest coil (Rapid Biomedical, Rimpar, Germany) was used for transmission and reception of 23Na and 1H signal. The built‐in body coil was used for 1H signal excitation. The operation frequency of the antenna is 32.6 MHz for 23Na nuclei and 123.2 MHz for protons. The coil is composed mainly from three modular components: an upper component (2 Tx 23Na, 2 Tx 1H), a flexible component (2 Rx 23Na, 2 Rx 1H) and a bottom component (2 Tx 23Na, 2ch Rx 23Na). The volume coil (top and bottom components) size is 57 (length) × 49 (width) × 37 (height) cm3 and the flexible receive coil size is 30 (length) × 35 (width) × 35 (height) cm3. The software for the pulse sequence allows us to operate the coil at either sodium or proton frequency.

Study participants

Written informed consent was obtained from six participants. The ethical board of the institution approved the study and all volunteers provided information prior to the examination. Due to macroscopic motion, data from one subject were rejected. Imaging with the same protocol was performed in the intervertebral disks of a total of five healthy volunteers (two males and three females) with an average age of 29.1 4.2 years.

Phantom measurements

To evaluate the distribution of B 1 field in a homogeneous medium, cylindrical phantom measurements were acquired. The phantom measurements were made using a volume cylindrical polyethylene container filled with 150 mmol/L NaCl. The same protocol and data processing as detailed below were applied. Additionally, five cylindrical phantom tubes filled with 50, 100, 150, 200 and 250 mmol/L NaCl with 5% agarose gel were used as a reference sodium signal intensity in experiments. The phantoms were located at the center of the volume coil and sodium concentration measurements were made with and without field correction to validate the applied 23Na imaging protocol and quantification framework.

MRI protocol

RF pulse calibration

A global RF pulse calibration was performed by adjusting the transmitter voltages between 200 and 340 V (15 measurements with linearly increasing nominal flip angle from 0° to 180°) using a nonlocalized rectangular 1 ms length RF pulse to ensure a true 90° flip angle at the center of the coil geometry before all measurements.

23Na‐MRI

Following a survey, sagittal and coronal 1H images were acquired for accurate positioning of the intervertebral disk. The spin‐density‐weighted 23Na images were acquired using a gradient echo (GRE) pulse sequence with the following parameters in supine position: T R/T E = 320/1.98 ms, FOV = 320 × 320 mm2, voxel size = 5 × 5 × 10 mm3, number of slices = 3, NEX = 128, nominal flip angle = 90°, total acquisition time = 5.56 min.

1H‐MRI

‐ and ‐weighted 1H anatomical imaging protocol was performed to facilitate image segmentation. The same coil was used for excitation and signal reception for 1H‐MRI without repositioning the subject. A FLASH sequence was used to acquire the ‐weighted structural scan (T R/T E = 308/4.77 ms, 24 interleaves, FOV = 320 × 320 mm2, voxel size = 1 × 1 × 1 mm3).

B 1 field mapping

The contrast variation of the 23Na‐MRI is fundamentally determined by the RF coils used for signal excitation and reception, and as expected the field patterns can drastically change based on the antenna involved. Sodium mapping was performed using the double‐angle method, which encodes the flip angle into the amplitude of the complex MRI signal. The field distribution is determined by the ratio of two images obtained at different nominal flip angles and . The field map was calculated based on the relationship between the effective and nominal flip angles from two GRE volumes acquired with 45° and 90° flip angles: where I 45 and I 90 are the intensities of the corresponding voxels in images acquired with 45° and 90° flip angles. A long repetition time (T R = 100 ms) was used to minimize the dependence in mapping. Note that only the transceive coil was considered, and based on the principle of reciprocity both transmit and reception field amplitudes are denoted as in this study. The flip angle maps were calculated on a voxel‐by‐voxel basis across the imaging volume. The correction factor was calculated for every voxel by dividing the actual flip angle obtained from the flip angle map by the nominal flip angle, and the resulting correction factor was applied to mitigate the inhomogeneity on the sodium images.

Sodium quantification

Sodium concentration quantification is usually done by placing reference phantoms with known sodium concentrations and relaxation times within the field of view of the images (i.e., the ranges of the reference tubes are usually 10 mM–150 mM to image brain or muscle, 100 mM–350 mM for cartilage). In the volunteer measurements, five standardized reference sodium tubes were included in the field of view to allow quantification of the TSC. Moreover, internal tissue references such as cerebrospinal fluid with a well defined sodium concentration of 150 mM/L were used as a reference and for validation.

Image processing and data analysis

Subsequent data processing and analysis were carried out using self‐developed scripts in MATLAB (MathWorks, Natick, MA, USA). TSC was calculated via an intensity calibration curve fitted to the signal of the reference tubes and the linear regression curve was used to extrapolate the sodium maps of the intervertebral disks. To minimize the influence of inhomogeneity on 23Na‐MRI, particularly in the specified geometry, all sodium‐weighted images were subjected to voxel‐wise ‐field correction. In vivo TSC values are presented as mean and standard deviation in millimoles per liter. To determine a region of interest (ROI) of intervertebral disks, ‐weighted 1H images were used. Intensity thresholding was performed after plotting the border contours, resulting in a binary mask, where individual intervertebral disks were identified. The identified binary masks defining the intervertebral disks were controlled by an external radiologist. The inhomogeneity corrected TSC maps were calculated with in the identified intervertebral disks.

In vivo quantitative TSC values for each subject are presented as mean and standard deviation in millimoles per liter. A paired‐sample t‐test was performed to compare the mean sodium concentrations of different subjects for the uncorrected and corrected cases.

RESULTS

The dual‐tunable (1H/23Na) chest coil used for reception of 1H, and for transmission and reception of 23Na signals, is shown in Figure 1A. Figure 1B shows a simulation of transmit field efficiency of the coil in the transverse axis illustrating that, while the field has relatively homogeneous pattern in the close vicinity of the transmit elements, it decreases towards the center of the coil. Figure 1C shows a spin‐density‐weighted sagittal image of the intervertebral disk geometry used as a localizer. The intensity variations on the image as a result of field inhomogeneity are observable. The five reference tubes were placed in the field of view with varying sodium concentrations. Figure 1D shows the flip angle map for proton signal, exhibiting a relatively high SNR in the intervertebral disks but poor SNR between the disks along the spine. The nonuniformity of the transmit field for protons is also visible on the reference tubes.

FIGURE 1

RF coil and proton field sensitivity. A, The dual‐tunable (1H/23Na) chest coil: the operation frequency of the 1H is 123.2 MHz and that for 23Na is 32.6 MHz. The coil is composed mainly from three modular components: an upper component (2 Tx 23Na, 2 Tx 1H), a flexible component (2 Rx 23Na, 2 Rx 1H) and a bottom component (2 Tx 23Na, 2ch Rx 23Na). The volume coil (top and bottom components) size is 57 (length) × 49 (width) × 37 (height) cm3 and the flexible receive coil size is 30 (length) × 35 (width) × 35 (height) cm3. B, Simulation of transmit field efficiency of the coil (V/m2) in the transverse axis, showing that, while the field has relatively homogeneous pattern in the close vicinity of the transmit elements, it decreases towards the center of the coil. C, Localizer image of the intervertebral disks obtained as a spin‐density weighted sagittal image. D, Flip angle map for proton signal using the body coil for transmission and the receiver elements of the rapid coil for reception exhibiting a relatively homogeneous transmit and receive field sensitivity with a relatively high SNR in the intervertebral disks but poor SNR between the disks

Figure 2 shows the effect of field in quantitative sodium imaging in a homogeneous phantom. Figure 2A shows the sodium concentrations of the reference tubes and the imaging phantom. Five reference tubes with 50 mmol/L, 100 mmol/L, 150 mmol/L, 200 mmol/L and 250 mmol/L NaCl with 5% agarose were placed above the imaging phantom with the sodium concentration of 150 mmol/L. Figure 2B shows the Na image of the phantoms with arbitrary units where the concentration gradient of the reference tubes from left to right is nicely depicted. Figure 2C shows the sodium flip angle map acquired with a double angle method (actual flip angle divided by the nominal flip angle). Even though the phantom and the reference tubes were placed at the center of the coil geometry and the imaging volume is much smaller compared with a human subject, there is up to 15% field nonuniformity across the phantom. The corresponding TSC maps with and without field correction are illustrated in Figure 2D and 2E respectively. The color bar refers to the actual sodium concentrations within the tubes and the phantom, demonstrating that the true concentrations were accurately reconstructed from the 23Na‐MR images. The effect of field correction is depicted in Figure 2F as the bar graphs of the field corrected and uncorrected concentrations within the tubes and the phantom. While the field exhibits relatively small variation within the imaging phantom, around 5%, the inhomogeneity increases along the reference tubes up to 15%, as implied by the flip angle map.

FIGURE 2

Effect of field in quantitative sodium imaging in a homogeneous phantom. A, Placement of the reference tubes with 50 mmol/L, 100 mmol/L, 150 mmol/L, 200 mmol/L and 250 mmol/L NaCl with 5% agarose and the imaging phantom with the sodium concentration of 150 mmol/L. B, Na image of the phantoms with arbitrary units where the concentration gradient of the reference tubes from left to right is depicted. C, Sodium flip angle map illustrated as the ratio of actual flip angle divided by the nominal flip angle. There is a field nonuniformity across the phantom up to 15% even though the phantom and the reference tubes were placed at the center of the coil geometry and the imaging volume is much smaller compared with a human subject. D, Quantitative TSC maps without field correction. E, Quantitative TSC maps with field correction. The color bar refers to the actual sodium concentrations within the tubes and the phantom, demonstrating that the true concentrations were accurately reconstructed from the 23Na‐MR images. F, bar graphs of the field corrected and uncorrected concentrations within the tubes and the phantom. While the field exhibits relatively small variation within the imaging phantom, around 5%, the inhomogeneity increases along the reference tubes up to 15% as implied by the flip angle map

Figure 3 shows the ROI placement using the ‐weighted 1H images. Figure 3A shows the ‐weighted image, where the individual intervertebral disk structures are clearly visible. Figure 3B shows the intensity thresholding of the ‐weighted structural image within the determined contours along the spine and the binary mask extracted from the thresholded images that is used as the ultimate ROI to calculate the TSC within the intervertebral disks. The intervertebral disks from T9/T10 to L5/S1 are labelled on the image. Figure 3C shows the ROI for CSF identified on the structural image as the internal tissue reference and validation of the sodium concentration measurement. Figure 3D illustrates the quantitative TSC map, illustrating that the sodium concentrations in the CSF were measured as 148 ± 7 within the defined ROI and the sodium concentrations within the intervertebral disks were accurately quantified based on the references.

FIGURE 3

ROI placement and TSC quantification. A, ‐weighted image of the intervertebral disks where the individual disk structures is clearly visible. B, Intensity thresholding of the ‐weighted structural image within the determined contours along the spine. The mask extracted from the thresholded images is used as the ultimate ROI. The intervertebral disks from T9/T10–L5/S1 are labelled. C, The ROI for CSF identified on the structural image. The Na concentration of the CSF was used as the internal tissue reference and validation of the sodium concentration measurements. D, The quantitative TSC map, illustrating that the sodium concentrations in the CSF were measured as 148 ± 7 within the defined ROI and the sodium concentrations within the intervertebral disks were accurately quantified based on the references

Figure 4 presents the effect of field in quantitative sodium imaging in an exemplary subject. Figure 4A shows 23Na images, where the sodium content of the intervertebral disks is clearly depicted in arbitrary units. The limited SNR provided by the utilized antenna prevents us imaging the whole spine structure at the same time. Figure 4B shows the in vivo sodium flip angle map depicted as the ratio of the actual to the nominal flip angle, which was deployed as the voxel‐by‐voxel correction factor. The magnetic field nonuniformity was measured up to 40% depicted by the flip angle map along and across the intervertebral disk geometry. Figure 4C and 4D shows the in vivo quantitative sodium concentration maps without and with field correction respectively. Even though a limited resolution was provided by the 23Na‐MRI, the anatomical structure of the intervertebral disks can be identified as annulus fibrosus and nucleus pulposus from the quantitative TSC maps. A further observation yielded by Figure 4D is that, even though the field variations were corrected, the quantitative TSC values are varying among the individual disks along the spine, which potentially signals physiological roots.

FIGURE 4

Effect of field in quantitative sodium imaging in an exemplary subject. A, 23Na images, where the sodium content of the intervertebral disks is clearly depicted in arbitrary units. The limited SNR provided by the utilized antenna prevents us imaging the whole spine structure at the same time. B, The in vivo sodium flip angle map depicted as the ratio of the actual to the nominal flip angle, which was further deployed as the voxel‐by‐voxel correction factor. The magnetic field nonuniformity was measured up to 40% depicted by the flip angle map along the intervertebral disk geometry. C, In vivo quantitative sodium concentration maps without field correction. D, In vivo quantitative sodium concentration maps with field correction. Even though a limited resolution was provided by the 23Na‐MRI, the anatomical structure of the intervertebral disks can be identified as annulus fibrosus (outer shell) and nucleus pulposus (inner shell) from the quantitative TSC maps. Even though the field variations were corrected, the quantitative TSC values are varying among the individual disks along the spine, potentially caused by physiological factors

The mean sodium concentration values are listed in Table 1 and Table 2 for uncorrected and corrected cases respectively for the individual disks. Depending on the disk position, the sodium concentration was calculated as 161.6 mmol/L–347 mmol/L and the mean sodium concentration of the intervertebral disks varies between 254.6 ± 54 mmol/L and 290.1 ± 39 mmol/L after correction. We observed a significant (p < 0.005) difference in the TSC after correction. Among individual disks of the same subject, the TSC values are significantly (p < 0.001) different compared with each other. In all subjects, there is a significant decrease in standard deviation of the mean sodium concentration with corection.

TABLE 1

Average quantitative sodium concentrations in the individual disks [mmol/L] calculated in the ROIs for each subject without B 1 field correction

IV disks	S1	S2	S3	S4	S5
T9/T10	136.5 ± 12	140.5 ± 15	134.6 ± 14	145.8 ± 16	148.4 ± 14
T10/T11	147.7 ± 18	207.2 ± 31	168.3 ± 19	197.2 ± 16	166.2 ± 20
T11/T12	193.8 ± 15	184.1 ± 24	211.2 ± 27	218.5 ± 27	207.4 ± 33
T12/L1	255.0 ± 33	276.5 ± 49	243.1 ± 28	257.5 ± 32	301.2 ± 41
L1/L2	298.0 ± 53	301.7 ± 39	324.0 ± 41	291.4 ± 49	316.2 ± 49
L2/L3	349.8 ± 63	321.3 ± 44	345.0 ± 51	306.5 ± 43	332.4 ± 52
L3/L4	344.4 ± 51	341.6 ± 52	319.8 ± 49	233.0 ± 34	253.2 ± 32
L4/L5	302.5 ± 42	268.8 ± 32	175.2 ± 21	198.5 ± 16	193.8 ± 26
L5/S1	174.9 ± 20	219.1 ± 28	150.2 ± 19	116.5 ± 10	145.2 ± 17
Mean ± std	244.7 ± 83	251.2 ± 67	230.1 ± 81	213.3 ± 62	229.3 ± 73

TABLE 2

Average quantitative sodium concentrations in the individual disks [mmol/L] calculated in the ROIs for each subject with B 1 field correction

IV disks	S1	S2	S3	S4	S5
T9/T10	199.5 ± 28	221.2 ± 29	204.7 ± 19	216.2 ± 17	200.7 ± 18
T10/T11	237.0 ± 21	268.8 ± 46	225.9 ± 22	234.1 ± 19	221.5 ± 21
T11/T12	245.1 ± 22	288.4 ± 48	250.2 ± 29	266.5 ± 37	246.1 ± 42
T12/L1	312.0 ± 48	304.5 ± 43	252.6 ± 30	288.5 ± 51	327.6 ± 51
L1/L2	293.8 ± 50	320.5 ± 45	300.0 ± 54	276.5 ± 46	334.8 ± 87
L2/L3	348.0 ± 57	347.0 ± 56	333.6 ± 59	340.6 ± 25	339.8 ± 56
L3/L4	337.9 ± 51	326.2 ± 43	328.2 ± 47	313.8 ± 32	260.4 ± 34
L4/L5	343.5 ± 43	288.4 ± 28	213.4 ± 28	197.5 ± 17	268.2 ± 36
L5/S1	231.3 ± 34	246.4 ± 30	183.2 ± 23	161.6 ± 13	259.2 ± 25
Mean ± std	283.1 ± 55	290.1 ± 39	254.6 ± 54	255.0 ± 57	268.4 ± 45

Figure 5 shows quantitative analysis of TSC within the individual disks and among different subjects. Figure 5A shows the variation of the average TSC values among the individual disks for a representative subject and the effect of field correction on it. While the effect of correction is relatively small for the disks positioned in the middle of the imaging volume (L1/L2, L2/L3, L3/L4), there is a prominent effect for the remaining disks located relatively close to the periphery of the imaging volume. Figure 5B shows the corrected TSC variation among the individual disks for different subjects. It is highly noticeable that the variation of TSC along the disks for different subjects provides similar patterns, yielding an increasing tendency from T9/T10 until L2/L3, where it peaks, and a decreasing pattern for the lower lumbar disks. Figure 5C shows the effect of field correction on the mean TSC of individual subjects (average of mean TSC for all disks for a single subject). The smoothing effect of correction by reducing the standard deviation in the average TSC for each subject is clear. Figure 5D shows the variation of normalized TSC (the ratio of TSC at each disk normalized by the mean TSC of all the disks for a single subject) for different subjects. As an analogy to coefficient of variation, the normalized TSC variation provides highly horizontal curves, implying a convergence to a constant ratio for every subject, varying between 0.6 and 1.3 for different disks.

FIGURE 5

Quantitative analysis of TSC within the individual disks and among different subjects. A, The variation of the average TSC values among the individual disks for a representative subject and the effect of field correction on it. B, corrected TSC variation among the individual disks for different subjects. It is highly noticeable that the variation of TSC along the disks for different subjects provides similar patterns, yielding an increasing tendency from T9/T10 until L2/L3 where it peaks and a decreasing pattern for the lower lumbar disks. C, Effect of field correction on the mean TSC of individual subjects (average of mean TSC for all disks for a single subject). D, Variation of normalized TSC (the ratio of TSC at each disk normalized by the mean TSC of all the disks for a single subject) for different subjects. As an analogy to coefficient of variation, the normalized TSC variation provides highly horizontal curves, implying a convergence to a constant ratio for every subject varying between 0.6 and 1.3 for different disks

DISCUSSION

In principle, in order to ensure an accurate quantitative TSC measurement, the acquired 23Na signal must be corrected for all contrast mechanisms apart from sodium concentration. The functional information extracted from the Na images can only be reliably applied if all signal modulations are well defined. The field at the sodium Larmor frequency emitted from a typical body‐volume antenna poses significant magnetic field inhomogeneity along the spine, directly affecting the spin evolution and ultimately the 23Na signal. In this study, we reported for the first time quantitative sodium concentrations in the intervertebral disks at clinical field strengths (3 T) by deploying 23Na‐MRI in healthy human subjects. This study is also the first to present sodium mapping in intervertebral disks and to assess the effect of sodium mapping to the quantitative TSC.

Dual resonant coils such as the one used in this work for sodium imaging are attractive mainly because they allow reference 1H images to be acquired without repositioning the subject. In contrast to quadrature birdcage coils, which usually provide relatively homogeneous excitation field patterns at 3 T, the body‐volume coils covering the intervertebral disk geometry exhibit highly inhomogeneous field profiles. Additionally, anatomical variations that differ in electrical permittivity from surrounding tissue further exacerbate the B 1 nonuniformities. The double‐angle method was used to calculate the flip angle maps due to easy practical implementation for sodium mapping. There are several other mapping techniques that could potentially be utilized such as actual flip‐angle imaging, Bloch–Siegert shift or the phase‐sensitive method. In comparison studies, it was previously reported that the Bloch–Siegert shift method and phase‐sensitive method are the most accurate methods for 1H. , For 23Na‐MRI, the phase‐sensitive method is shown to yield higher quality maps at low signal‐to‐noise ratio and greater consistency of measurement than the double‐angle method, but with higher vulnerability to large off‐resonance shifts. Recently, a method for simultaneous mapping and imaging was proposed in order to enhance accuracy and to reduce measurement time with higher SNR compared with the double‐angle method, which could potentially be applied.

To mitigate problems associated with low signal levels obtained during in vivo sodium imaging and sodium mapping experiments, the image resolution was limited to pixel sizes of a few millimeters (5 × 5 × 10 mm3). However, such a voxel volume causes the accuracy of 23Na‐MRI to be strongly biased by partial volume effects (PVEs). Previously reported partial volume correction (PVC) methods for 23Na‐MRI for the brain tissue could be adapted for the spinal cord geometry.

Phantom measurements were conducted to evaluate the quality of the Na‐weighted images and mapping. The experiments showed that the correction could significantly improve the quantitative accuracy of the sodium concentration maps besides smoothing the excitation profile, implied by the reduced standard deviation. Thus we performed correction for the in vivo data. In previous work concerning the 23Na‐MRI in the head using a transceive birdcage coil, correction had up to 10% difference in TSC for white matter, cerebrospinal fluid and vitreous humour. The larger influence of the correction on the measured TSC values that was observed in our work is caused by the higher nonuniformity profile of the body‐volume coil used. The sodium concentration values of the intervertebral disk of 161.6 mmol/L–347.0 mmol/L agree with previous studies using sodium MRS. , , After the application of mitigation, the standard deviation of the mean sodium concentration was significantly reduced (p < 0.001) and the smoothing effect of the correction is obvious.

23Na‐MRI is a promising quantitative method that might be helpful to determine the biochemical status of the human intervertebral disk and better understand the pathophysiology of disk degeneration. Previous studies have reported that a ‐weighted semiquantitative grading system and Pfirrmann classification can be used for the assessment of healthy and degenerative disks. , In the clinical routine, measurement of sodium representing the PG content of the disk might provide a noninvasive tool for assessment of disk degeneration at an early stage, giving the possibility for prevention and to monitor treatment effects, providing a quantitative tool. The quantitative sodium concentration in the intervertebral disks at clinical field strengths provided in this work with correction of the transmit field variations further improves the accuracy of the overall clinical 23Na‐MRI framework.

This study has some limitations concerning the technical aspects. First of all, no correction was performed. Although the maps we obtained from the phantom measurements exhibit negligible off‐resonance (below 3 Hz), susceptibility artifacts are highly prevalent along the spinal cord due to different susceptibilities in the intervertebral disks, and the lungs cause significant variations in the field. However, the shimming is highly challenging in the spinal cord geometry using the standard shim procedure mainly because not all inhomogeneity can be compensated with only first‐ and second‐order shim coils. Higher order shimming and slice‐wise dynamic shimming are needed for sufficient compensation. Moreover, the dynamic variation of the field due to the respiration cycle and moving dielectric tissue such as chest and abdomen can cause significant field fluctuations that require dynamic higher order shimming. It has been previously reported that in vivo 23Na nucleus exhibits a fast biexponential transversal relaxation time such that 60% of the signal decays with a of 0.7–3 ms (fast component) and 40% of the signal decays with a of 16–20 ms (slow component). Consequently, it is important to acquire images at a short T E to avoid the signal loss from the fast‐relaxing component. We acquired images at 1.98 ms, leading to a residual weighting of the signal that can be further reduced by using ultra‐short echo time acquisition schemes. , Regarding this study, limitations include the small number of participants as well as missing reproducibility assessment. Both tasks will be taken as future work, particularly to study the short‐ and mid‐term variability of the sodium concentration measurement in the intervertebral disks, which is most interest for clinical studies to monitor variations due to underlying physiological dynamics.

CONCLUSION

In conclusion, quantitative 23Na‐MRI is a promising tool to measure clinically relevant longitudinal changes in the intervertebral disks. Here, we reported quantitative sodium concentration measurement in the intervertebral disks at clinical field strengths (3 T). We showed that the nonuniformities associated with the antenna utilized for imaging significantly modulate the ultimate TSC and have to be mitigated for quantitative sodium concentration mapping. The results of this work have the potential to enable an improved integration of quantitative 23Na‐MRI into clinical studies in intervertebral disks such as degenerative disk disease and establish alternative scoring schemes to existing morphological scoring such as the Pfirrmann score.