In Vivo 1H MR Spectroscopy of Biliary Components of Human Gallbladder at 7T

Background Previous in vivo proton MR spectroscopy (MRS) studies have demonstrated the possibility of quantifying amide groups of conjugated bile acids (NHCBA), olefinic lipids and cholesterol (OLC), choline‐containing phospholipids (CCPLs), taurine and glycine conjugated bile acids (TCBA, GCBA), methylene group of lipids (ML), and methyl groups of bile acids, lipids, and cholesterol (BALC1.0, BALC0.9, and TBAC) in the gallbladder, which may be useful for the study of cholestatic diseases and cholangiopathies. However, these studies were performed at 1.5T and 3T, and higher magnetic fields may offer improved spectral resolution and signal intensity. Purpose To develop a method for gallbladder MRS at 7T. Study Type Retrospective, technical development. Population Ten healthy subjects (five males and five females), two patients with primary biliary cholangitis (PBC) (one male and one female), and one patient with primary sclerosing cholangitis (PSC) (female). Field Strength/Sequence Free‐breathing single‐voxel MRS with a modified stimulated echo acquisition mode (STEAM) sequence at 7T. Assessment Postprocessing was based on the T2 relaxation of water in the gallbladder and in the liver. Concentrations of biliary components were calculated using water signal. All data were corrected for T2 relaxation times measured in healthy subjects. Statistical Tests The range of T2 relaxation time and concentration per bile component, and the resulting mean and standard deviation, were calculated. Results The concentrations of gallbladder components in healthy subjects were: NHCBA: 93 ± 66 mM, OLC: 154 ± 124 mM, CCPL: 42 ± 17 mM, TCBA: 48 ± 35 mM, GCBA: 67 ± 32 mM, ML: 740 ± 391 mM, BALC1.0: 175 ± 92 mM, BALC0.9: 260 ± 138 mM, and TBAC: 153 ± 90 mM. Mean concentrations of all bile components were found to be lower in patients. Data Conclusion This work provides a protocol for designing future MRS investigations of the bile system in vivo. Evidence Level 2 Technical Efficacy Stage 1

substrates, such as bilirubin and bile salts. Bile salts are the major organic solutes in bile and normally function to emulsify dietary fats and facilitate their absorption. Bile is also the major route for elimination of cholesterol and protects the organism from enteric infections, and some biliary compounds (eg, bile salts, bilirubin) undergo an enterohepatic circulation. 1 Changes in bile composition are not only relevant for gallstone formation but may also play an important role in the pathogenesis and treatment of chronic inflammatory bile duct diseases. 2 Importantly, the stimulation of phospholipid secretion may be a key therapeutic mechanism of several established and new drugs for treatment of cholangiopathies such as primary biliary cholangitis (PBC) and primary sclerosing cholangitis (PSC). 2 Bile can be characterized in vitro in the micellar phase with proton nuclear magnetic resonance spectroscopy (MRS) analysis. [3][4][5][6] Pioneer work in gallbladder in vivo MRS in humans was published by Prescot et al, 7 where cholinecontaining phospholipid signal (CCPL, 3.22 ppm) at a magnetic field strength (B 0 ) of 1.5T was detected. Künnecke et al showed reliable detection of several other bile components in monkeys at 4.7T, 8 including taurine and glycine conjugated bile acids (TCBA, 3.08 ppm, and GCBA, 3.74 ppm, respectively). More recent MRS studies of healthy subjects and patients with PSC were performed at 3T. 9,10 Ultrahigh field of 7T has promising advantages for MRS, such as high spectral resolution and higher signal-tonoise ratio (SNR), which could help better resolve bile components and acquire spectra faster.
Relaxation correction, with longitudinal (T 1 ) and transversal (T 2 ) relaxation times, is necessary for the calculation of accurate bile component quantification by MRS. T 2 relaxation times of bile components were reported in vitro from porcine bile 9,11,12 and from bile of dogs. 13 There is a need for assessment of human bile relaxation times as well. Studies in humans employed point-resolved spectroscopy sequence (PRESS) 14 with echo times (TE) of 30 msec and 60 msec. 7,9 These TEs are suitable for measurements of longer relaxing bile components, such as CCPL; however, a sequence with a TE < 30 msec would be advantageous for assessment of the short relaxing components. A modified ultrashort-TE stimulated echo acquisition mode (STEAM) sequence has been used for liver MRS at 7T. 15 Since the T 2 relaxation time of water in gallbladder is several fold longer than the T 2 relaxation time of water in the liver, 9,16 analyzing the water linewidth in water unsuppressed spectra could provide unique and definitive information about the origin of the signal and guide signal discrimination during postprocessing.
The aims of this study were to establish a measurement and postprocessing protocol for in vivo measurement of concentrations of human bile components corrected for T 2 relaxation times measured in healthy subjects with ultrashort TE MRS during free breathing without water suppression at 7T and to apply the protocol in patients with cholangitis.

Human Subjects
The study protocol was approved by the Institutional Review Board and the Ethics Committee, and written, informed consent was obtained from all participants prior to enrollment in the study. In vivo MRS measurements were performed on 10 healthy volunteers (age 33.5 ± 4.1 years, body mass index [BMI] 22.4 ± 1.3 kg.m -2 , five males and five females), two patients with primary biliary cholangitis (PBC) (age 47 and 63 years, BMI 24.7 kg.m -2 and 31.2 kg. m -2 , one male and one female), and one patient with primary sclerosing cholangitis (PSC) (age 47 years, BMI 22.0 kg.m -2 , female).
The healthy subjects had no history of gallbladder disease. The subjects had no history of acute, chronic liver disease or inflammatory bowel disease. None of the subjects reported any special dietary restrictions.

In Vivo Gallbladder MRS
All MRS experiments were performed with a 7T Magnetom MR system (Siemens Healthineers, Erlangen, Germany) with a doubletuned 1 H/ 31 P surface coil ( 1 H double loop 14 × 23 cm, Rapid Biomedical, Rimpar, Germany). All subjects were scanned in the supine position after overnight fasting. Fasting was confirmed by imaging a full gallbladder on the T 1 -weighted scout images. The center of the coil was placed over the lowest ribs, on the right side of the rib cage.
The spectroscopic voxel was positioned according to T 1weighted scout images (gradient echo, relaxation time / echo time [TR/TE] = 7.8/3.7 msec, field of view = 400 × 400 mm 2 , slice thickness = 6 mm). The frequency was adjusted on the water signal, and the homogeneity of the B 0 field was first improved with an automatic B 0 shimming method calculated from gradient echo images. Linear shims were then manually adjusted in an interactive mode. The accepted water linewidth was 60 Hz or less in magnitude mode.
For MRS, a STEAM sequence with hermite pulses (bandwidth of 3.57 kHz) and spoiler gradient switching 15 was used. The signal was measured with mixing time (TM) of 10 msec and acquired with 1024 complex points, and the receiver bandwidth was set to 3000 Hz. TR was set to 5 seconds. No outer volume suppression was used. The reference voltage for RF pulses was adjusted according to the water signal amplitude by incrementally changing the voltage of the RF pulses from 280-340 V. More information about the sequence, RF pulses, and their adjustment can be found elsewhere. 15 Due to the narrow frequency profile of hermite pulses and chemical shift displacement error (CSDE), an acquisition with a frequency offset set on water (0.0 ppm) was used for quantification of signals from 8 ppm to 3 ppm, and a separate acquisition with an offset of -3.4 ppm from water resonance was used for quantification of signals from 3 ppm to 0 ppm. The size and the position of the voxel were individually adjusted according to the size of the gallbladder, with the size ranging from 1.0-3.4 cm 3 (mL). The linewidth of the water signal was observed on a real-time display. The localization of a voxel was considered to be accurate when a series of relatively narrow water peaks was detected. The number of spectral transients per frequency offset (NT) was 8. Spectral transients were stored individually.
Every healthy subject was first measured with a TE of 6 msec, 50 msec, 100 msec, and 150 msec. Additional measurements with TE of 20 msec, 30 msec, 70 msec, 120 msec, and 200 msec were performed, according to subject compliance. Patients were measured only with a TE of 6 msec. The total acquisition time per one TE was 1.33 minutes. Total examination time including patient positioning, shimming, and sequence adjustments ranged from 15-20 minutes.

Data Postprocessing
Signals were processed offline with in-house software written in MATLAB (MathWorks, Natick, MA). A Fourier transform was applied to all eight individual spectral transients per TE and frequency offset, the full-width at half-maximum of water (FWHM H2O ) was calculated from magnitude spectra, and spectra with FWHM H2O <60 Hz (equivalent of 30 Hz in absorption mode) were selected for further signal processing. The FWHM H2O limit of 60 Hz in magnitude mode was used for processing of all signals. The selected transients were automatically phased, frequency aligned, and averaged. The resulting spectrum was truncated (last 200 complex points) and zero-filled to 4096 complex points in jMRUI. 17 The chemical shifts of bile components were obtained from the literature 3, 18  The signals were fitted with Lorentzian line shapes using the AMARES algorithm. 19 Because no water suppression was used during the data acquisition, the water signal was removed offline with the Hankel-Lanczos singular value decomposition (HL-SVD) method 20 in jMRUI. This method was successfully applied in the in vivo MRS study at 7T. 15 In case the water sidebands were not efficiently removed, the HL-SVD method was used to remove them at a chemical shift of ±553 Hz (±1.9 ppm) from the water peak.

T 2 Relaxation Times
Monoexponential T 2 relaxation was assumed for all bile components. Individual T 2 relaxation times were calculated in MATLAB with the following monoexponential fitting function: where S 0 represents the signal intensity at TE = 0 msec. The coefficient of determination (R 2 ) was used as an indicator of the quality of the fit. Mean T 2 relaxation times of bile components were calculated from the individual T 2 relaxation times from all subjects. The calculation of T 2 relaxation times are described in more detail in the Supplemental Material (S2). The T 2 relaxation times are reported as mean ± standard deviation and range.

Concentrations of Bile Components
Concentrations in mM were calculated according to Mohajeri et al and Fayad et al 9,21 : where I BC,T2 and I H2O,T2 are the corrected signal intensities for individual T 2 relaxation times for bile components and water, respectively; NP BC and NP H2O are the number of protons for bile component and water (NP H2O = 2), respectively; M H2O is the molar mass of water (≈18.015 g/mol); and CF H2O is a conversion factor for the water content of human gallbladder bile (CF H2O = 0.9). The concentrations are reported as mean ± standard deviation and range.

Results
In Vivo MRS An example of surface coil position and placement of voxels and gallbladder MRS are shown in Fig. 1a,b. Mean size of the voxel in healthy subjects was 2.9 ± 1.0 mL. This limit allowed keeping on average at least 50% of the transients with average FWHM H2O from 15.0 Hz to 20.2 Hz. An overview of the number of subjects measured per specific TE, mean number of accepted scans per frequency offset and TE, and FWHM H2O per frequency offsets and TE, is summarized in Tables S1 and S2 in the Supplemental Material. An example of the bile spectrum measured at 7T is depicted in Fig. 2a. An example of spectra used for the calculation of T 2 relaxation times for bile components is depicted in Fig. 2b. Mean spectra of all subjects measured with TE of 6 msec are depicted with their respective standard deviations in Fig. 2c,d.
T 2 Relaxation Times T 2 relaxation times were calculated for all 10 bile components in all subjects except for NHCBA and OLC in one subject, due to low SNR. The T 2 relaxation times of all bile components and water signals with corresponding R 2 values are summarized in Table 1. Comparisons of previously published T 2 relaxation times measured in vitro to T 2 relaxation times measured in this study are shown in Table 2.

Concentration of Bile Components
The lowest mean concentration in healthy subjects was found in the case of CCPL, 42 ± 17 mM (range: 19-66 mM). The concentration of CCPL had the smallest standard deviation of all bile components. Other relatively low concentrated metabolites were TCBA and GCBA. Comparable concentrations were found in the case of TBAC, OLC, and BALC1.0. The highest concentrations were in the case of BALC0.9 and ML. mean concentrations of bile components with respective standard deviations and ranges measured in healthy subjects and patients are summarized in Table 3. Comparison of previously published concentrations of previously measured bile components in healthy subjects and in patients with subjects in this study is shown in Table 4.

Discussion
In the work presented here, the concentration and T 2 relaxation rates of bile components were measured with in vivo MRS at 7T using a modified STEAM sequence with free breathing and an ultrashort TE without water suppression in 10 healthy subjects and three patients with cholangitis. The high spectral resolution of 7T enabled detection and quantification of water and other nine bile components. The detection of GCBA, TCBA, and methyl signals of bile acids, lipids, and cholesterol was superior to 3T. Although prone positioning for gallbladder MRS has been proposed to minimize respiratory movements and susceptibility differences at tissue-air interfaces, 22 the supine position used in our study provided data with sufficient quality. The mean size of the spectroscopic voxel in our study was similar to the mean voxel size reported by Prescot et al (3.3 mL). 7 Acquisitions with breath-hold and respiratory gating techniques are feasible and widely used for liver and gallbladder MRS, 9,23 yet there are several factors to consider. The placement of the volume of interest (VOI) depends on anatomical images acquired during the same respiratory phase. Even then, there is no guarantee that the prescribed VOI will a b c d be in the same location, since the subject's movement during the breathing phase could change the position of the VOI relative to the investigated tissue. This problem is more pronounced in the gallbladder than in the liver due to the nature of its small size and range of shapes. 24 Previous approaches with MRS employed triggering techniques, 8,9 which enabled acquisition of the gallbladder spectra without contamination and high SNR. The triggering techniques use an effective TR that is adjusted to the breathing cycle of a subject. Our approach fixed the TR to 5 seconds, the average time for a breathing cycle. 9 The choice of using a TR of 5 seconds was to minimize any possible T 1 effects as well. According to measurements performed on porcine bile at 3T, 9,11,12 the T 1 relaxation time of water was 780 msec and the longest T 1 relaxation time of other bile components was below 400 msec. Abdominal MRS employing surface coils at 7T is particularly challenging due to the positioning, B 1 power constraints and inhomogeneity, 25 difference in relaxation times, 26,27 and respiratory motions. The modified STEAM sequence for 7T has been shown to be a reliable method for the measurement of hepatic lipids. 15 We applied the same method for the gallbladder. Narrow spectral bandwidths and frequency ranges of RF pulses in this sequence needed to be compensated by two measurements with different frequency offsets. 15 Although these two measurements doubled the acquisition time, this approach also minimized the CSDE, which can be substantial at 7T. 15 For example, in the case of the 3.4 mL cubical voxel (1.5 × 1.5 × 1.5 cm 3 ) and the RF pulse centered at the water frequency, the CSDE for CCPL was only 12.3%, but for NHCBA it was 27.4%. In the case the pulse was centered at ML, the CSDE for TBAC was only 4.9%. The weakness of quantification of the NHCBA signal is discussed below.
Due to short TEs and strong spoiler gradients used in this sequence, unsuppressed strong water signals cause the occurrence of water sidebands. The sidebands result from vibrations of the gradients 28 and/or oscillation of the  The T 2 of water and CCPL was measured at 1.5T in porcine bile 7 ; the T 2 of water was measured at 2.35T in canine bile 13 ; and the T 2 of water and the other five bile components was measured at 3.0T in porcine bile. 9,11,12 magnetic field. 29 In an ideal case, the spoiler orientation switching method 30 eliminates the sidebands, but only when pairs of transients with opposite spoiler gradient orientations are acquired. For example, a water sideband at 2.80 ppm may appear close to the TCBA peak (3.08 ppm) in the gallbladder spectra. The HL-SVD method proved to be sufficient to remove the water sidebands in the case of the matching pairs of transients were not available. Postprocessing of the data required an efficient FWHM limit for the water signal. Due to the significant differences in T 2 relaxations of water in the liver 15 and the gallbladder (12 msec vs. 145 msec, respectively), the water linewidth in the gallbladder spectra was considerably narrower. Although we did not measure B 0 distribution in the abdomen, this difference seemed to be more significant than differences in B 0 homogeneity. An MR cholangiopancreatography (MRCP) is based on heavily T 2 -weighted pulse sequences, taking advantage of the inherent differences in T 2 relaxations between stationary fluid-filled structures in the abdomen (long T 2 relaxation time) and neighboring soft tissue (short T 2 relaxation time). 31 The same principle was applied in our proposed MRS acquisition. The FWHM H2O limit used in this study of 30 Hz showed to be sufficient for acquisition of data from gallbladder. Importantly, it worked for both acquisitions with different frequency offsets, where the number of accepted transients and FWHM H2O were comparable. Variations of water amplitude were observed across all subjects. These individual differences in the water amplitude can be attributed to different positions of the gallbladder relative to the surface coil; various shapes of the gallbladder; the voxel size and its placement. We used an NT of 8 per offset, which effectively gives an NT of 16 for acquisition of all signals. To minimize these variations and get higher SNR, we recommend increasing the NT to 32. This allows for better selection of the spectral transients based on their FWHM H2O . The acquisition time in this study was shorter than in previous studies:  3.2 minutes at 1.5T 7 and 22.89 minutes at 3T. 9 The preparation time included in the total examination time can be beneficial for other measurements; for instance, abdominal MRI and 1 H MRS of the liver. These measurements can be performed in the same session with the same setup, except for voxel adjustments in the case of MRS. T 2 relaxation times of bile components were previously measured in vitro only in animals. 9,11,13,32 We did not observe a large decrease of the T 2 relaxation time of water at 7T compared to literature values, 9,11,13,32 which suggests mostly dipolar interactions of water molecules without a large contribution of diffusion and susceptibility effects. 27 The exception is the T 2 relaxation time of water in porcine bile measured at 1.5T published by Prescot et al. 7 We did not include a possibility for J-modulation of bile components in the calculations of T 2 relaxation times for two reasons. First, because spin-spin J-coupling interactions in bile components measured with STEAM have not yet been reported, so we decided to use the simpler approach for their assessment. Second, the high precision for assessment of T 2 relaxations times was of less importance for quantification because we used very short TE. The mean coefficients of determination of all components were higher than 0.7, which suggests acceptable assessment of T 2 relaxation with monoexponential decay. However, additional studies are needed to establish a standard range of T 2 relaxation times of bile components in human gallbladder in vivo in both healthy subjects and in patients.
The concentration of CCPL is in agreement with previous in vivo studies in humans. 7,9 Although we reported the concentration of NHCBA, it must be noted that the chemical shift difference from water of this signal is 3.3 ppm, which at 7T represents substantial chemical shift displacement error (CSDE). Moreover, the frequency profile of the RF pulse will underestimate this signal, as described previously. 15 Despite this suboptimal acquisition, this signal could still be detected and quantified because of the small voxel size, which resulted in a VOI that was within the gallbladder even for bile components with substantial CSDE.
The concentrations of all bile components had large variations across all healthy volunteers. Variations have been observed in vitro in bile 33 and in vivo MRS in monkeys 8 and in humans. 9 Khan et al 34 implied that these variations may be due to variations in diet, nutrition, use of drugs, and even exposure to different environmental factors that affect bile secretion and gallbladder motility. A greater number of subjects with controlled dietary environments and restrictions are needed to better characterize the variability of bile components. Future studies will need to focus on the reproducibility of the results as well. Moreover, the composition of hepatic (bile duct) vs. gallbladder bile may be of primary interest for the pathogenesis and treatment of bile duct diseases.
Lower concentrations of bile components in patients with PSC have already been reported. 10 The authors of that study suggested that the reduction in bile acid secretion could indicate accumulation of bile acids in hepatocytes. The results reported in our study support previous findings. Changes in concentrations of bile salts and phospholipids and their ratios can increase bile salt toxicity and, thereby, contribute to bile duct injury, as seen in a range of hereditary and acquired bile duct diseases. 35 Importantly, the stimulation of phospholipid secretion may be a key therapeutic mechanism of several established and new drugs such ursodeoxycholic acid, obeticholic acid (as well as other farnesoid X receptor agonists), and fibrates for treatment of cholangiopathies such as PBC and PSC. 2 Moreover, impaired phospholipid excretion has been linked to a broad spectrum of hereditary and acquired hepatobiliary disorders, 35 including nonanastomotic structures after liver transplantation and low phospholipidassociated cholelithiasis syndrome. Notably, stimulation of biliary phospholipid excretion may be an important therapeutic mechanism in new drugs used for the treatment of cholangiopathies. 2 Detection and quantification of phosphatidylcholine in the gallbladder and liver can be alternatively performed using phosphorus MRS imaging ( 31 P MRSI) at 7T. 36,37 Phosphatidylcholine protects bile ducts from the harmful effects of bile acids by the formation of mixed micelles. 38 The combination of MRS with 31 P MRSI at 7T could provide an even more detailed picture of the gallbladder.

Limitations
Detecting abdominal motions are one of the main limitations of this study. Our setup did not contain any gating technique to reduce motion artifacts. Although our approach was able to filter out signals that were not gallbladder-specific, a combination of techniques for motion detection could increase the number of accepted transients per unit time. Future studies should focus on incorporating these techniques. B 1 field of the surface coil is another limitation of this study. The variation of signal amplitude was largely dependent on the anatomy of the subject: distribution and thickness of the adipose tissue and size and location of the gallbladder. Although the surface coil used in this study had an exceptional penetration metric, the position of the gallbladder could still be on the edge of reasonable signal detection. A coil array design for body MRI 39 could be a solution for imaging the abdominal cavity and measuring the gallbladder using MRS at UHF in the future.
In some cases, due to CSDE, the NHCBA signal could be detected from different tissues or not at all. The large variations of this signal can be explained by different anatomical variations of the gallbladder. This is a weakness of our protocol. In the case that NHCBA is of high interest, the pulse needs to be centered on the NHCBA signal (offset = +3.3 ppm).
The number of healthy subjects and patients was relatively small. This work provided a basis for future gallbladder MRS experiments and the relatively small sample sizes served as proof of principles and not as a reference for concentrations of bile components.

Conclusion
In vivo 1 H MR spectroscopy of the gallbladder during free breathing without water suppression is feasible at 7T. A short acquisition time and ultrashort TE enabled measurements of T 2 relaxation times of 10 bile components, including water. The methods used in this study may allow for in vivo measurement of bile components; however, additional work needs to be done to establish standard ranges and the repeatability of the method. Importantly, this work provided a protocol for designing future MRS investigations of the bile system in vivo.