• high field;
  • liver;
  • MRI


  1. Top of page
  2. Abstract
  3. Basic aspects of magnetic resonance spectroscopic methodology
  4. Quantification of metabolites
  5. The magnetic resonance spectrum
  6. Clinical applications and future directions
  7. References

With increased availability of magnetic resonance (MR) systems at ultra-high field strength for clinical studies, other organs besides the brain have received renewed consideration for MR spectroscopy (MRS). Because signal-to-noise ratio and chemical shift increase proportional to the static magnetic field, a concomitant increase in signal intensity and spectral resolution of metabolite resonances can be exploited. Improved resolution of adjacent metabolite peaks would not only provide for more accuracy of metabolite identification but also metabolite quantification. While the superiority of high-field imaging and spectroscopy has already been demonstrated clearly in the brain, this article reviewed issues around 1H MRS of the liver. These include optimization strategies such as coil technology, minimizing of motion artefacts using breath-holding and postprocessing of the spectra. Moreover, we reviewed the pertinent experience hitherto reported in the literature on potential clinical issues where liver MRS may be useful. These included determination and characterization of liver fat content, liver tumours and focal lesions. While these applications have been used experimentally, liver MRS does not yet have a clearly defined role in the clinical management of any disease state. Accordingly, it remains primarily a research modality to date.

The clinical use of localized 1H magnetic resonance spectroscopy (1H MRS) in vivo first in the brain and then in the prostate has been well established and refined over the last two decades (1–5). Proton spectroscopy in other organs also has a similar long history in experimental studies, but has always been a challenge in clinical environments owing to signal distortions in moving organs, lack of diagnostic resonances and clinical non-practicability. With increased availability of MR systems at ultra-high field strength for clinical studies, other organs besides the brain have received renewed attention. Because signal intensity and chemical shift increase proportional to the static magnetic field B0, a concomitant increase in signal-to-noise ratio (SNR) and spectral resolution of metabolite resonances is expected.

The ability to resolve different adjacent metabolite peaks better would not only provide for more accuracy of metabolite identification but also metabolite quantification. Hence, the conclusion as to which peaks are abnormal would be facilitated (6, 7).

In vitro MRS of tissue extracts, which are mostly performed of fields of >11.7 T (500 MHz) today, allows a glimpse at the potentially detectable metabolic tissue components in vivo and may itself provide for a comprehensive metabolic profile of steady-state concentrations in endogenous pathways (8–10). MRS in vivo is more difficult to perform because of non-ideal conditions including the need for gradient shimming owing to field inhomogeneities and the need for water suppression and localization with appropriate pulse sequences. It is characterized by a much poorer spectral resolution and SNR than in vitro MRS of homogenous solutions in spinning test tubes. Factors such as the lower magnetic field strengths used for in vivo clinical studies, magnetic susceptibilities and patient motion contribute to relative spectral degradation. On the other hand, increasing the field strength for in vivo studies to those of in vitro nuclear magnetic resonance will introduce a number of complicating factors such as hardware constraints, increase of distorting susceptibility effects and T2 relaxation time shortening. This would suggest a field strength optimum for clinical human MRS in the range of 3.0–7.0 T, dependent on the objective and the organ under study as well as technical factors.

Doubling the field strength from 1.5 to 3.0 T results in a two-fold gain in SNR, as the increase in SNR is directly proportional to the field strength (11). Thus, smaller voxels may become possible at 3.0 T compared with 1.5 T or, otherwise, the number of excitations may be decreased to allow for more acquisition speed. This consideration, of course, is only valid when the radio frequency (RF) coils used are comparable. It should be noted that improvements in RF coil configuration, e.g. with phased-array multichannel technology, can easily make up for a doubled SNR as well.

Out of several other MRS-sensitive nuclei such as carbon (13C), fluorine (19F) or phosphorus (31P), especially the latter has been extensively studied already in the ‘preproton era’ in the early and mid-1980s as information about liver metabolism, intracellular pH and the energy status could be gained despite problems with proper localization (12). 1H MRS, however, has the most valuable advantage that the same hardware for magnetic resonance imaging (MRI) as well as for MRS can be used. Moreover, its sensitivity is by a factor of 7 higher than that of phosphorus, so that volumes of interest (VOIs) of typically 4–8 cm3 have been applied for 1H MRS instead of 30–100 cm3 for 31P MRS. Other nuclei are not only less abundant or sensitive than the proton, but always need dedicated coil systems tuned to the specific Larmor frequency of the nucleus at the desired field strength (13).

The human liver seems to be technically attractive for 1H spectroscopy, because the organ is large and readily accessible. Modern multi-element surface receiver coils operated in a phased array can be placed easily around the abdomen.

In addition, the liver is an attractive organ clinically as metabolic changes such as fatty infiltration can be a substantial part of therapy monitoring. Also, preoperative diagnostic MRI of small hepatic metastases or of hepatocellular carcinoma (HCC) in cirrhotic liver may still be challenging. Based on a single MRI examination, it is still very difficult to differentiate malignant from benign regenerating nodules in liver parenchyma (14). Hence, a reliable and objective technique would be highly desirable to identify the pathology unequivocally.

Basic aspects of magnetic resonance spectroscopic methodology

  1. Top of page
  2. Abstract
  3. Basic aspects of magnetic resonance spectroscopic methodology
  4. Quantification of metabolites
  5. The magnetic resonance spectrum
  6. Clinical applications and future directions
  7. References

Magnetic resonance spectroscopy is a valuable tool for the non-invasive assessment of metabolic processes in vivo. Neuronal markers, membrane constituents, osmolytes and the energy status can be measured for the diagnosis of various diseases and therapeutic monitoring in humans (3).

1H magnetic resonance spectroscopy generates a spectrum of the various resonances of protons that are embedded in different chemical bonds. Because the protons are surrounded by various nuclei and electrons with their own magnetic properties, small magnetic field pertubations occur in a systematic manner, leading to slight differences in the received frequencies of protons in different chemical bonds. Thus, the chemical shift values occur essentially as a consequence of the variable electronegativity of adjacent chemical moieities in the molecule. The chemical shift scale describes the position of resonances in the spectrum in parts per million (p.p.m.), irrespective of the field strength, relative to a reference set at 0 p.p.m. (Si(CH3)4). The underlying frequency shift, however, measured in Hertz (Hz), is directly proportional to the strength of the magnetic field, e.g. 1 p.p.m. of the proton spectrum at 1.5 T refers to 64 Hz and at 3.0 T to 128 Hz. Therefore, with higher magnetic fields the resonances are better separated. The frequency separation of the resonances or peaks describes the resolution of the spectrum.

As the same nuclei are responsible for MRI and MRS in case of the proton, anatomical imaging can be carried out before spectral analysis in order to determine the exact position of the voxel. In case of a focal hepatic lesion, the voxel is then graphically prescribed within the bulk of the tumour as identified in the images using a pulse sequence for localization.

Single volume spectroscopy with stimulated echo acquisition mode or the point-resolved spectroscopy sequence (PRESS) technique is recommended (15, 16), because of longer acquisition times and reduced SNR for multivoxel liver MRS with chemical shift imaging (15–17).

To achieve a maximum signal, at first, the choice of high field strength and RF coil configuration will be of prime importance. Second, parameter choices including a high number of signal averages and a large volume will increase the signal. Of note, T1-weighted or T2-weighted spectra can largely be avoided by long repetition times and short echo times, respectively, facilitating quantification (2, 17).

In order to obtain the highest SNR possible for MRS acquisitions, dedicated multi-element coil arrays would be desirable. The use of 32 elements as a receiver phased-array coil has been demonstrated to yield significantly superior SNR compared with the whole-body resonator or even four-element or eight-element receiver coils, providing the ground for high-quality spectra in a single breath-hold (18). The modern multichannel phased-array coils are well tolerated by the patients and provide for a quite homogenous B1 field. If these preferred coil systems are not available, then the whole-body resonator might also be used both for RF signal transmission and reception. This leads to a decrease in the SNR and longer acquisitions times, but when applying appropriate acquisition protocols phased-array coils might not be required (19). In addition, the whole-body resonator increases homogeneity in the magnetic field significantly, which is crucial for appropriate shimming.

Using the PRESS technique in the liver, the repetition time should exceed 2 s with 128 averages in order to obtain enough signal from a minimum VOI of 2 × 2 × 2 cm3 (8 ml using the preferred coils discussed above). The SNR or spectral quality of smaller voxels is generally considered not to be sufficient, despite the use of a 3.0 T high-field system.

The choice of short echo times decreases T2 signal losses, in particular, of metabolites with relatively short T2 relaxation times such as lipids. The preset flip angles used, initially optimized for brain tissue using simulations with known T1 and T2 relaxation times of the water, were also found to be applicable for liver spectroscopy.

So far, the use of abdominal MRS has lagged behind that in the central nervous system owing to local B0 field inhomogeneities and artefacts caused by motion such as respiratory movement, vessel pulsation, the beating heart and intestinal peristalsis.

In fact, one of the main concerns is the ability to deal with motion artefacts. The effects of motion on MRS include voxel misregistration leading to outer-voxel contamination, which may or may not be apparent in the spectra. Phase and frequency shifts that are caused by movement of the tissue through inhomogenous B0 and B1 fields result in broadening of the spectral resonances. Moreover, increased blood and iron contents of the liver compared with the brain increase the spectral water linewidth at full-width at half-maximum to >12 Hz depending on the size of the VOI, homogeneity shim and field strength.

Therefore, hepatic MRS can be challenging. Several techniques have been used to reduce motion artefacts including the navigator pulse technique, respiratory gating or signal averaging. None of these methods eliminates the degradation of spectral quality entirely. Most satisfactory results are obtained with breath-hold acquisitions, although this technique may not be well tolerated by all patients and imposes a patient-dependent time limit on the acquisition (20). The total acquisition time must be split into consecutive blocks to match the length of a breath-hold period of about 20–40 s, while data acquisition is performed at end-expiration.

Nevertheless, especially measurements in the upper parts of the liver, directly under the diaphragm and near the apex of the heart, remain most susceptible to motion.

Besides the ability to deal with motion artefacts, the diagnostic value of abdominal MRS relies on adequate technical factors such as the prescan adjustments and effective water suppression.

For MR systems suffering from very strong eddy current artefacts, it may become necessary to improve spectra by phase cycling of the excitation pulses. This is a commonly used method to reduce baseline artefacts by averaging a series of acquisitions at different phases, which usually results in averaging out a significant amount of artefacts. The remaining motion artefacts, though, can compromise data averaging over several cycles. Therefore, phase cycling of only two cycles can be superior to the conventional eight cycles, which are familiar in brain spectroscopy.

Because the average concentration of water in normal tissue is several magnitudes higher than the concentrations of the MR-detectable metabolites at 0.5–10 mM, water suppression has to be performed (1). Commonly, chemical shift-selective suppression pulses are used.

As the efficiency of water suppression also suffers from distortions owing to motion, additional postprocessing steps are required to take the varying amount of water suppression into account.

Retrospective image averaging (21, 22) is not a standard software on conventional scanners, but improves SNR and allows for increased acquisition time without introducing motion artefacts.

Further processing steps include phase and frequency correction as phase regularization (22), exponential filters, which emphasize the signal-bearing part of the free induction decay, interpolation of the data points with zero-filling and correction of eddy currents.

For quantitative analysis, the full water signal should be available at a concentration reference. This requires an additional acquisition without water suppression, which adds to the examination time.

Although this procedure is crucial for optimal spectral analysis, there are still many variable facets and, so far, it is not described in a uniform manner in the manuals.

Quantification of metabolites

  1. Top of page
  2. Abstract
  3. Basic aspects of magnetic resonance spectroscopic methodology
  4. Quantification of metabolites
  5. The magnetic resonance spectrum
  6. Clinical applications and future directions
  7. References

In principle, the area under the resonance signal of a specific chemical residue that represents the metabolite is directly related to the concentration of the residue-bearing metabolite. However, absolute quantification of metabolites is difficult to achieve in vivo. Until today, absolute quantification remains a challenge (23–26). Therefore, it is recommended that each site establishes its own reference concentration values for comparison.

Field inhomogeneities owing to interferences and dielectric resonance effects at high field strengths prevent the use of the reciprocity theorem for a direct estimate of metabolite concentrations based on the measured signal intensity (23–26). Nevertheless, a technically robust approach of using phased-array coils at 2.9 T for a quantification of absolute metabolite concentrations has been proposed using an eight-element receive-only head coil for single-voxel localized proton MRS of human brain. The method is based on the transmitter reference amplitude of the body coil used for RF transmission (23). A relative sensitivity of every element of the phased array coil is derived from a combination of two reference scans without water suppression that correspond to either the body coil in transmit–receive mode or the phased array coil in conjunction with body coil excitation.

Numerous absolute quantification techniques have been proposed for spectroscopy of the liver (27). These techniques involve the calibration of in vivo signals from a VOI by comparison with an internal or an external reference of known concentrations, in addition to knowledge of the relaxation parameters of the metabolites of interest.

For an external reference, a spectrum is recorded with identical acquisition parameters from a voxel of identical size placed in a phantom containing the reference compound at a known concentration. The phantom is placed next to the abdomen. The spectra are acquired separately and the results are compared to determine the concentration in the tissue.

Unfortunately, variations in coil geometry and varying flip angles over the area covered must be taken into account. Values might be distorted because of dielectric resonance effects and field inhomogeneities (24). Differences in T1 and T2 relaxation times between the phantom and the tissue should also be considered. The external reference procedure requires extra scanning time in two separate measurements and delicate calibration, which may not be feasible in a clinical context, particularly when the spectroscopy examination is additionally prescribed onto a standard clinical imaging protocol. In addition, patients also complained about discomfort relating to the use of an external water bottle that was placed next to their body to investigate loading effects.

Nevertheless, Li et al. (27) performed experiments at 3.0 T using an external phantom containing choline chloride for calibration. They claimed the method to be accurate and requiring fewer tedious procedures, although various investigations for T1 and T2 times and system stability had to be performed.

The signal used for an internal reference stems from known endogenous compounds that have a defined concentration. Metabolite concentrations can then be reported in terms of metabolite ratios under the assumption of constancy of the reference compound. Obviously, this assumption may not be given in pathological states of the liver.

Also, the number of available metabolites in liver spectroscopy, which can be quantified reliably, is very limited. In contrast to the muscle and brain, where the creatine peak is often used as a reference, signals from creatine do not appear in spectra of the liver, because hepatocytes do not express creatine kinase under normal circumstances (28).

The use of the lipid peak as an internal reference should not be considered, as the range of lipids is very variable, changing with the body mass index (BMI), weight and sex. Moreover, there may be a non-uniform distribution of the lipids owing to a heterogenous distribution of steatosis. Recent surveys suggest that about 25% of the population suffer from liver steatosis with an increased risk owing to obesity, alcohol overuse, hyperlipidaemia and diabetes mellitus (29).

Thus, the only remaining reference and a simple and practical approach in the clinical environment for application is the use of the unsuppressed water signal. If all the metabolites that contribute to the choline-containing compounds' (CCC) peak can be considered to be water soluble, then the internal unsuppressed water can be used as an internal standard to measure CCC concentrations.

It still needs to be taken into account, however, that in addition to pathological situations, there are physiological conditions that affect the water content, e.g. it is still unclear so far how fasting or time of day affects MRS.

Li et al. (27) reported a 1.8-fold difference between the largest and smallest water signal intensity obtained from the localized liver tissues.

The problem mentioned is very similar to the problem of ‘quantitative’ breast spectroscopy, where currently, several groups use water as an internal reference reasonably successfully owing to a lack of alternatives.

The magnetic resonance spectrum

  1. Top of page
  2. Abstract
  3. Basic aspects of magnetic resonance spectroscopic methodology
  4. Quantification of metabolites
  5. The magnetic resonance spectrum
  6. Clinical applications and future directions
  7. References

In a liver proton spectrum, independent of the field strength between 1.5 and 3.0 T, four resonances are usually detected including the dominant lipids between 0.8 and 2.3 p.p.m., with the methylene proton signals [(–CH2–)n and (–CH2–CH=CH–)] at 1.2 and 2.1–2.3 p.p.m., respectively, and the methyl proton signal (CH3–) at 0.9 p.p.m., and the trimethylammonium resonances, made up of betaine and CCC, at 3.2 p.p.m. (see Fig. 1). Unresolved resonances at 3.35–3.9 p.p.m., which probably originate from myo-inositol, glycogen and glucose moieties, γ-CH2– of glutamine and glutamate at 2.2–2.6 p.p.m., from methine residues (–CH=CH–) at 5.5 p.p.m., and the previously described unknown resonance at 4.0 p.p.m. are detected very inconsistently (8). As the SNR for these resonances is low in vivo, it is difficult to differentiate as to which are genuine resonances and which are noise-related signals. Hence, these peaks cannot be used for diagnostic purposes.


Figure 1.  PRESS-localized single voxel 1H MR spectrum (TR/TE=2000/35 ms, NEX=128, vol=8 ml) originating from healthy liver parenchyma of a young volunteer. The location of the spectroscopic VOI is indicated by the square in the T2W image insert displaying the right liver lobe of a normal liver. The spectral calculation reveals resonances in accordance with the Methyl (CH3–, 0.9 ppm) and Methylen protons (–CH2–CH2–, 1.2 ppm; –CH=CH–CH2–, 2.0–2.3 ppm) present in fatty acids or lipids. The resonance at 3.2 ppm arises from the Trimethylammonium (TMA) and/or Choline-containing compounds (CCC).

Download figure to PowerPoint

Evaluation of hepatic triglyceride content

Fatty infiltration of the liver is a common condition that represents a nonspecific response to many metabolic and or toxic disorders.

While the lipids in the hepatic cell membrane should not be altered by fatty infiltration or other hepatic disorders and do not contribute to the in vivo proton MR signal, the prominent lipid resonances allow a good estimate of the hepatic fat content.

Various MRI techniques have been proposed for the detection or visualization of liver steatosis. The fat content in the regions of interest of images acquired with fat-selective spectral excitation as described by Machann et al. (30) correlated well with spectroscopic results (r>0.95).

T1-weighted images show a relatively bright signal of fatty liver owing to the short relaxation time T1 of triglycerides. However, most human studies have reported that the T1 and T2 relaxation times in fatty livers range within the normal values (31).

The Dixon technique is based on the chemical shift differences between water and fat resonances owing to their different Larmor frequencies. Water- and fat-selective images can be obtained by adding and subtracting the in-phase and opposed-phase images (32). However, the sensitivity to low amounts of hepatic lipids is potentially limited.

A series of detailed studies have validated MRS measures of hepatic fat content against values obtained by computed tomography (CT), histomorphometry and biochemical extraction (33, 34).

Longo et al. (33) described a method for quantifying fat infiltration of the liver using 1H MRS studies of lipid extracts from steatotic liver specimens by determining the lipid volume fraction with the definition of a ratio of the fat signal peak area (FTSA) of the detectable fat signal peak area (region 0.5–3.0 p.p.m.) to the total signal peak area (region 0.5–5.5 p.p.m.). The signals from lipid protons resonating between 3.0 and 5.5 p.p.m. were not considered because of ‘masking’ by the water proton resonance.

Assuming a ratio of 0.85 between the signal integrals in the two spectral regions, the lipid volume fraction can be expressed by the formula (33)

  • image

A comparison of the results showed that MRS slightly overestimated fat infiltration compared with CT (<4%), while histomorphometry based on biopsy specimens underestimated the fat content of liver by about 30 and 50%.

Using the equations validated by Longo et al. (33), Szczepaniak et al. (35) used 1H MRS to analyse the distribution of the hepatic triglyceride content in 2349 patients from the Dallas Heart Study. The study demonstrated that 1H MRS can be used well in a clinical setting to assess hepatic fat content.

Intake of a high-fat meal did not significantly affect the measurements, and the values were similar when measurements were made from the right and left hepatic lobes. The upper limit of the normal level for hepatic triglycerides was determined by examination of 345 subjects who had no identifiable risk factors for hepatic steatosis (nonobese, nondiabetic, minimal alcohol consumption, normal liver function tests and no known liver disease).

The 95th percentile in these subjects was 5.56%, which corresponded to a hepatic triglyceride level of 55.6 mg/g. This value corresponds to the traditional threshold of 50 mg/g (5% by wet weight) that is considered to be diagnostic of hepatic steatosis (34, 36). Based on this cut-off value, the prevalence of hepatic steatosis in that study was estimated to be 33.6%.

Machann et al. (30) also revealed a marked interindividual variability of hepatic lipid levels in the range of 0.5–39.3%. Moreover, their study showed a homogenous fat distribution in individual livers with no tendency towards the presence of different amounts of lipids in distinct subregions or segments of the liver.

In contrast, Tarasow et al. (37) reported a variation of 6.5–87.4% in the fat content from two different positions in the liver, although only five subjects underwent examinations of two different voxel regions out of the 24 healthy volunteers investigated.

Thomas et al. (38) stated that there appeared to be no significant regional differences in the relative levels of fat distribution, but substantial interindividual variation. Some volunteers showed a difference in fat content of up to 50% between regions, whereas in others it was <1%.

Mahmood et al. (39) observed a statistically significant positive correlation between the hepatic lipid concentration and the BMI, suggesting that the tendency towards obesity is also connected to lipid accumulation in the liver. The relation of the hepatic fat content to the body adiposity has been confirmed by Thomas et al. (38), who showed that the fat signals detected in all subjects were significantly larger in overweight subjects compared with lean subjects (with triglyceride contents >5.5%).

Also, Schonfeld et al. (40) found an increased fat content in obese patients compared with controls, although the latter still showed considerable levels of hepatic triglycerides.

On the other hand, no significant relationship was observed in a group of 20 volunteers between the thickness of subcutaneous fat tissue and lipid concentration in the liver (36).

Mahmood et al. (39) reported a negative correlation between the visceral and subcutaneous fat tissue ratio in men and the degree of liver attenuation in CT images.

All studies were performed at a lower field strength of 1.5 T. In our own in vivo studies, at 3.0 T, a wide range of hepatic fat content was observed in the study population, even though there was no history of liver disease (41). Lower levels were generally found in younger and healthier volunteers, while older less active subjects showed a higher content.

It is still unclear to date as to which factors determine and modulate accumulation of hepatic fat in human subjects, both in health and in disease. A genetic susceptibility appears to be likely. A thorough understanding of the possible heterogeneity of liver triglyceride distribution is important, because it may have major implications for interventional studies.

More recently, the relative quantification of intrahepatocellular lipid content has been used successfully to evaluate the pathogenesis of conditions such as insulin resistance and type II diabetes. Seppala-Lindroos et al. (42) showed that the fat accumulation in the liver, as determined by 1H MRS, was associated with defects in the insulin suppression of glucose production and serum-free fatty acids. No dependence on obesity was observed in normal men, as determined by MRI. The study showed that the ability of insulin to suppress serum-free fatty acids was impaired in those with high liver fat compared with those who had low liver fat, independent of the intra-abdominal fat and body weight. The study supports the view that fat accumulation in tissue is an important determinant of its sensitivity to insulin. This is also supported by the Third National Health and Nutrition Examination Survey, showing that adults with non-alcoholic fatty liver disease were more than twice as likely to have diabetes than those without fatty liver (35).

The theory that hepatic insulin resistance is a consequence of increased hepatic fat is currently no more than a hypothesis because of lack of experimental support. In this respect, the reduction of increased hepatic fat might be a new therapeutic target.


A further valuable use of MRS has been shown by Cho et al. (43), who proposed to use proton MRS as a method for staging liver cirrhosis. With liver cirrhosis, the risk of developing HCC is elevated (44). Indirect evaluation parameters such as derived from ultrasound, CT and MRI are being used widely in conjunction with biochemical serum markers to assess the status of chronic liver disease (45). However, these methods were not sufficient for the correct staging of hepatitis.

In the Ludwig classification system, chronic hepatitis is classified into five histopathological stages on the basis of the severity of fibrosis subsequent to necroinflammatory insults (46).

Cho et al. (43) examined 75 patients with chronic hepatitis and measured the relative metabolite-to-lipid ratios by dividing the peak areas of the glutamine and glutamate complex, glycogen and glucose complex and the CCC with that of the lipids. These ratios were correlated with histopathological features. The most significant spectral change found at chronic hepatitis was a decrease in the lipid signal. When the stage of chronic hepatitis became higher, all the metabolite-to-lipid ratios increased. Explanations for the results are a decrease in the amount of lipid in the liver parenchyma or a relative increase in the amount of the other metabolites, when the stage of chronic hepatitis becomes higher (43). In addition, an increase of metal content such as of iron or copper may cause local magnetic field inhomogeneities that result in a decrease of the lipid signal.

Although these results appear to be promising, in our opinion, they have to be treated with caution. Because the amount of lipids varies and the SNR in the reported spectra appears to be quite low, it is difficult to discern as to which are genuine resonances and which are noise-related signal variations.


A further addendum to liver spectroscopy might be the investigation of hepatic bile to identify and quantify constituents. Bile consists of a variety of dissolved metabolites including cholesterol, esters, phospholipids, bile acids, fats, glycoproteins, various salts and xenobiotic substances (47). The complexity of bile contents is increased by the fact that it varies between and within patients, depending on the nutritional status and drugs. Diseases will affect the bile composition and the composition of gallbladder bile remains an important determinant of biliary lithogenicity (48).

Several in vitro studies have been carried out to investigate the quantitative determination and distribution of biliary lipids between vesicles and micelles, which is believed to have a role in gallstone diseases (49).

As phospholipid is a significant component of bile, 31P MRS studies may be particularly more relevant to the study of bile, because the technique can define phospholipids very readily. Nevertheless, Khan et al. (50) investigated proton and phosphorus MRS of human bile samples in hepatopancreaticobiliary cancer using 11.0 T. This ex vivo study revealed a reduced phosphatidylcholine signal in the majority of cancer patients compared with that in non-cancer patients in both 1H and 31P MRS. These preliminary studies suggest that spectroscopy of bile may be used to detect differences in phospholipid contents in cancer patients when bile samples can be collected in patients who have undergone endoscopic retrograde cholangiopancreatography.

Also, the in vivo spectra of bile have been shown to mainly display the phosphatidylcholine resonance at 3.2 p.p.m. Further metabolite resonances are substantially attenuated owing to a limit to shorter echo times leading to substantial peak suppression of J-coupled spins and coincident chemical shifts (49).

Although in vivo 1H MRS is technically feasible for the assessment of the human gallbladder, the method is currently of limited clinical value. This is because cholesterol signals have not been detected and macroscopic motion is a source of peak attenuation and contamination with surrounding tissue, especially in a poorly distended gallbladder (51, 52).

Evaluation of focal hepatic lesions

In addition to the lipid resonances, the 1H MRS of liver exhibits the CCC peak at 3.2 p.p.m. An improved differentiation of the resonance contributing to the CCC can be derived from 31P spectroscopy, which shows that the choline resonance includes contributions from phosphodiesters (PDE) such as glycero-3-phosphocholine (GPC) and glycero-3-phosphoethanolamine (GPE), and phosphomonoesters (PME) such as phosphocholine (PC) and phosphoethanolamine (PE). PME and PDE denote characteristic signals found at particular chemical shifts. These signals share a common chemical feature of phosphomonoester or phosphodiester bonds. The contributions of choline and acetylcholine to the resonance are negligible. There are also no contributions from bound membrane phospholipids, as the spectra arise from mobile soluble metabolites only (53).

Elevated PME contents represent a general observation for most tumour cells and tissues and are responsible for the increased choline resonance observed in MRS of cancer lesions in vivo. The PME are known to be part of the phospholipid synthesis (54). The increased levels have been hypothesized to be associated with intensified cell membrane synthesis and rate of cell replication (55). An extensive body of literature suggested that PME signals could be a possible diagnostic marker for tumours. According to Bell et al. (8), tumour tissue contains significantly elevated levels of PME and significantly lower concentrations of PDE than healthy liver tissue. PDE are suggested to be the breakdown products of phospholipids (56), and the concentration may be an indicator of the necrotic fraction in tumours associated with phospholipid catabolism (57).

1H magnetic resonance spectroscopy in vivo at 1.5–7.0 T does not allow to resolve the multiplets of these resonances observed in vitro. For the proton spectrum, the main signal contribution has been assigned to GPC besides trimethylammonium resonances including betaine and other CCC (9).

An ex vivo 1H MRS study on liver tissues at 1.5 T has shown an elevated resonance at 3.2 p.p.m. and decreased lipids in HCC compared with cirrhosis and normal liver, suggesting that quantitative 1H MRS in vivo could provide clinically useful information (58).

Kuo et al. (59) reported 1H MRS in vivo to be technically feasible also at 3.0 T for the evaluation of focal hepatic lesions and noted limitations in distinguishing between normal liver, benign and malignant tumours.

Rectal adenocarcinoma, which often metastasizes into the liver, was reported by Dzik-Jurasz et al. (60) to show proton spectra at 3.0 T dominated by resonances of CCC in vivo.

However, in vivo proton MRS of focal hepatic lesions has only been reported by a few studies so far. Kuo et al. (59) included altogether 33 lesions (21 HCC, two angiosarcomas, one lymphoma and nine haemangiomas) and compared these tumours with uninvolved liver tissue and, in addition, with the healthy livers of eight volunteers. In this study, the mean CCC/lipid ratios in different groups and after transcatheter arterial chemoembolization of eight patients were measured, showing metabolite changes after chemoembolization. No significant differences between normal liver and malignant tumours were observed. While the technical success rate for MRS was stated to be 90%, they did not use a breath-hold technique to exclude movement-related artefacts.

Li et al. (27) assessed the usefulness of applying an external phantom replacement method to measure the Cho concentration in four cases of HCC. They performed a comparison with the Cho concentrations in the healthy liver of five young adults acquiring the spectra under free breathing conditions. While the intensities varied significantly between individual volunteers, they reported substantially higher values for the tumours.

Soper et al. (58) used in vitro 1H MRS to characterize liver biopsy samples into normal, cirrhotic or HCC on the basis of a computer-based statistical classification strategy. Changes in lipids, choline and creatine were identified. Eighty-nine per cent of HCCs in this series were distinguished from non-malignant tissue on the basis of reduced lipid and increased choline contents.

Our results suggest only a tendency towards increased CCC levels in spectra of HCC lesions (41). Overall, malignant entities did not show elevated CCC levels compared with normal liver. On the contrary, rather a decrease was observed, specifically for metastases of rectal and breast cancer.

Because a high variability was observed, a large number of cases are considered to be necessary for further studies. With few exceptions, studies have included fewer than 50 study subjects. Small sample sizes make firm conclusions about the utility difficult.

The reasons for these apparent discrepancies might be that the concentrations of PC and PE vary considerably between tumours (61–63). Moreover, alterations of the metabolites are not exclusive to malignancy and also occur in developing organs as well as in proliferating healthy tissues. This might help to explain why most prominent choline resonances have been found in lean young adults.

Another reason for underestimating the CCC resonance in the malignant tumour group may be the relatively poor SNR that still remains with a higher field strength of 3.0 T, which compromises separation of the CCC from the noise level. In contrast to the healthy control group that performed breath-hold easily, the malignant group sometimes did not tolerate the breath-hold approach well, owing to the clinical condition of the patients.

It should also be pointed out that lesions might contain microscopic necrotic areas, cysts or haemorrhage, not evident on MRI. Also, the tumour is often surrounded by oedema and it is not always evident where the tumour ends and oedema begins.

The MR-visible metabolite changes within the viable part and the necrotic part of the tumour may differ. Owing to the relatively large voxel size, a partial volume effect with the necrotic portion of the tumour is likely to occur in several cases. As necrotic areas show a scarcity of metabolites, they would dilute the more prominent changes observed in areas of rapid cell turnover encountered in viable tumour tissue.

It is also notable that some signal contribution may arise from non-liver cells such as endothelial cells. Hepatocytes comprise approximately 60% of the liver by cell number and 80% of the overall volume (64).

So far, a differentiation between tumour and liver parenchyma is not straightforward, as normal liver may already contain large choline metabolite pools as we have shown. While catabolic and anabolic reactions are predominant in the liver, leading to elevated PDE and PME levels, it is very difficult to differentiate normal parenchyma from cancer. The situation in other organs such as muscle or the female breast appears to be different because of little or no CCC in the normal spectrum (65, 66).

Clinical applications and future directions

  1. Top of page
  2. Abstract
  3. Basic aspects of magnetic resonance spectroscopic methodology
  4. Quantification of metabolites
  5. The magnetic resonance spectrum
  6. Clinical applications and future directions
  7. References

There are many potential clinical applications for hepatic 1H MRS, especially for high field systems equipped with phased-array coil technology that provide improved SNR and spectral resolution.

The method has the potential to replace liver biopsy in the diagnosis of diffuse fatty infiltrations and might be used as a follow-up for the differentiation between reversible and irreversible alterations according to the level of fatty infiltration. In addition, smaller or greater volumes can be measured with MRS at the same or different locations in order to estimate temporal or focal alterations, e.g. in tumour therapy.

Already, the method has been instrumental for studies investigating the mechanisms underlying hepatic insulin resistance, as the accumulation of fat in insulin-sensitive tissue is a valuable parameter independent of the obesity of the patient (38, 64).

Nevertheless, liver MRS does not yet have a clearly defined role in the clinical management of any disease state and, accordingly, remains primarily a research modality to date.

Indeed, the various fat content levels measured in volunteers that were similar to or higher than those in patients with biopsy-proven steatosis suggest that those individuals represent ‘the tip of the iceberg’ in epidemiological terms.

Therefore, further work will be required to ascertain the ‘normal’ level of the internal hepatic fat content in the general population and establish the levels at which triglyceride deposition becomes pathological.

In combination with facilitated use of acquisition and processing protocols, abdominal spectroscopy may become a valuable method that can answer questions beyond the differential diagnosis of focal liver lesions. Foreseeable applications focus on the diagnosis and follow-up of degenerative liver diseases.


  1. Top of page
  2. Abstract
  3. Basic aspects of magnetic resonance spectroscopic methodology
  4. Quantification of metabolites
  5. The magnetic resonance spectrum
  6. Clinical applications and future directions
  7. References
  • 1
    Bruhn H, Frahm J, Gyngell ML, et al. Noninvasive differentiation of tumors using localized 1H MR spectroscopy in vivo: initial experience in patients with cerebral tumors. Radiology 1989; 172: 5418.
  • 2
    Michaelis T, Merboldt KD, Bruhn H, Hänicke W, Frahm J. Absolute concentrations of metabolites in the adult human brain in vivo. Quantification of localized proton MR spectra. Radiology 1993; 187: 21927.
  • 3
    Ross B, Bluml S. Magnetic resonance spectroscopy of the human brain. Anat Rec 2001; 2: 5484.
  • 4
    Cunningham CH, Vigneron DB, Marjanska M, et al. Sequence design for magnetic resonance spectroscopic imaging of prostate cancer at 3 T. Magn Reson Med 2005; 535: 10339.
  • 5
    Horn JJ, Coakley FV, Simko JP, et al. Prostate cancer: endorectal MR imaging and MR spectroscopic imaging – distinction of true-positive results from chance-detected lesions. Radiology 2006; 238: 1929.
  • 6
    Gruetter R, Weisdorf SA, Rajanayagan V, et al. Resolution improvements in in vivo 1H NMR spectra with increased magnetic field strength. J Magn Reson 1998; 135: 2604.
  • 7
    Barker PB, Hearshen DO, Boska MD. Single-voxel proton MRS of the human brain at 1.5 T and 3.0 T. Magn Reson Med 2001; 5: 7659.
  • 8
    Bell JD, Cox IJ, Sargentoni J, et al. A 31P and 1H-NMR investigation in vitro of normal and abnormal human liver. Biochim Biophys Acta 1993; 1225: 717.
  • 9
    Podo F. Tumour phospholipid metabolism. NMR Biomed 1999; 12: 41339.
  • 10
    Khan SA, Cox IJ, Hamilton G, Thomas HC, Taylor-Robinson SD. In vivo and in vitro nuclear magnetic resonance spectroscopy as a tool for investigating hepatobiliary disease: a review of H and P MRS applications. Liver Int 2005; 2: 27381.
  • 11
    Edelstein WA, Glover GH, Hardy CJ, Redington RW. The intrinsic signal-to-noise ratio in NMR imaging. Magn Reson Med 1986; 3: 60418.
  • 12
    Wehrli FW. Principles of magnetic resonance. In: StarkDD, BradleyWG, eds. Magnetic Resonance Imaging, Chapter 1. St Louis: Mosby, 1999; pp 114.
  • 13
    Solga SF, Horska A, Clark JM, Diehl AM. Hepatic 31P magnetic resonance spectroscopy: a hepatologist's user guide. Liver Int 2005; 3: 490500.
  • 14
    Jeong YY, Mitchell DG, Kamishima T. Small (<20 mm) enhancing hepatic nodules seen on arterial phase MR imaging of the cirrhotic liver: clinical implications. Am J Roentgenol 2002; 178: 132734.
  • 15
    Frahm J, Merboldt KD, Hänicke W. Localized proton spectroscopy using stimulated echos. J Magn Reson 1987; 72: 5028.
  • 16
    Bottomley PA. Selective volume method for performing localized NMR spectroscopy. US patent No. 4480228, 1984.
  • 17
    Salibi N, Brown MA. Clinical MR Spectroscopy. First principles. New York: Wiley-Liss, 1998.
  • 18
    Schirmer T, Sodickson DK, Rofsky NM, et al. 1H liver spectroscopy with 32-element coil arrays. Proc Int Soc Magn Reson Med 2005; 13: 1929.
  • 19
    Schirmer T, Thormann M, Fischbach F, Winter L, Bruhn H, Ricke J. Semi-quantitative liver spectroscopy with a transmit/receive body resonator at 3 T. Proc Int Soc Magn Reson 2004; 1: 924.
  • 20
    Earls JP, Rofsky NM, Decorato DR, Krinsky GA, Weinreb JC. Echo-train STIR MRI of the liver: comparison of breath-hold and non-breath-hold imaging strategies. J Magn Reson Imaging 1999; 9: 8792.
  • 21
    Feinberg DA, Rofsky NM, Johnson G. Multiple breath-hold averaging (MBA) method for increased SNR in abdominal MRI. Magn Reson Med 1995; 6: 9059.
  • 22
    Star-Lack JM, Adalsteinsson E, Gold GE, Ikeda DM, Spielman DM. Motion correction and lipid suppression for 1H magnetic resonance spectroscopy. Magn Reson Med 2000; 3: 32530.
  • 23
    Natt O, Bezkorovaynyy V, Michaelis T, Frahm J. Use of phased array coils for a determination of absolute metabolite concentrations. Magn Reson Med 2005; 53: 38.
  • 24
    Collins CM, Liu W, Schreiber W, Yang QX, Smith MB. Central brightening due to constructive interference with, without, and despite dielectric resonance. J Magn Reson Imaging 2005; 21: 1926.
  • 25
    Schirmer T, Auer DP. On the reliability of quantitative clinical magnetic resonance spectroscopy of the human brain. NMR Biomed 2000; 1: 2836.
  • 26
    Danielsen ER, Michaelis T, Ross BD. Three methods of calibration in quantitative proton MR spectroscopy. Magn Reson B 1995; 3: 28791.
  • 27
    Li CW, Kuo YC, Chen CY, et al. Quantification of choline compounds in human hepatic tumors by proton MR spectroscopy at 3 T. Magn Reson Med 2005; 53: 7706.
  • 28
    Cox IJ. Development and applications of in vivo clinical magnetic resonance spectroscopy. Prog Biophys Mol Biol 1996; 65: 4581.
  • 29
    Lonardo A, Bellini M, Tartoni P, Tondelli E. The bright liver syndrome. Prevalence and determinants of a “bright” liver echopattern. Ital J Gastroenterol Hepatol 1997; 4: 3516.
  • 30
    Machann J, Thamer C, Schnoedt B, et al. Hepatic lipid accumulation in healthy subjects: a comparative study using spectral fat-selective MRI and volume-localized 1H-MR spectroscopy. Magn Reson Med 2006; 55: 9137.
  • 31
    Stark DD, Goldberg HI, Moss AA, Bass NM. Chronic liver disease: evaluation by magnetic resonance. Radiology 1984; 150: 14951.
  • 32
    Dixon WT. Simple proton spectroscopic imaging. Radiology 1984; 153: 18994.
  • 33
    Longo R, Pollesello P, Ricci C. Proton MR spectroscopy in quantitative in vivo determination of fat content in human liver steatosis. J Magn Reson Imaging 1995; 5: 2815.
  • 34
    Thomsen C, Becker U, Winkler K. Quantification of liver fat using magnetic resonance spectroscopy. Magn Reson Imaging 1994; 12: 48795.
  • 35
    Szczepaniak LS, Nurenberg P, Leonard D, et al. Magnetic resonance spectroscopy to measure hepatic triglyceride content: prevalence of hepatic steatosis in the general population. Am J Physiol Endocrinol Metab 2005; 288: 4628.
  • 36
    Hoyumpa AM Jr, Greene HL, Dunn GD, Schenker S. Fatty liver: biochemical and clinical considerations. Am J Dig Dis 1975; 20: 114270.
  • 37
    Tarasow E, Siergiejczyk L, Panasiuk A, et al. MR proton spectroscopy in liver examinations of healthy individuals in vivo. Med Sci Monit 2002; 8: 3640.
  • 38
    Thomas EL, Hamilton G, Patel N, et al. Hepatic triglyceride content and its relation to body adiposity: a magnetic resonance imaging and proton magnetic resonance spectroscopy study. Gut 2005; 54: 1227.
  • 39
    Mahmood S, Taketa K, Imai K, et al. Association of fatty liver with increased ratio of visceral to subcutaneous adipose tissue in obese men. Acta Med Okayama 1998; 52: 22531.
  • 40
    Schonfeld G, Patterson BW, Yablonskiy DA. Fatty liver in familial hypobetalipoproteinemia: triglyceride assembly into VLDL particles is affected by the extent of hepatic steatosis. J Lipid Res 2003; 44: 4708.
  • 41
    Thormann M, Schirmer T, Fischbach F, Bruhn H, Felix R, Ricke J. Quantitative in-vivo 1H-MR-spectroscopy of liver lesions and liver parenchyma at 3.0 T. Proceedings of the RSNA: #SSE10-03, 2004.
  • 42
    Seppala-Lindroos A, Vehkavaara S, Hakkinen AM. Fat accumulation in the liver is associated with defects in insulin suppression of glucose production and serum free fatty acids independent of obesity in normal men. J Clin Endocrinol Metab 2002; 87: 30238.
  • 43
    Cho SG, Kim MY, Kim HJ, et al. Chronic hepatitis: in vivo proton MR spectroscopic evaluation of the liver and correlation with histopathologic findings. Radiology 2001; 221: 7406.
  • 44
    Saunders JB, Walters JRF, David P, Paton A. A 20-year prospective study of cirrhosis. Br Med J 1981; 282: 2636.
  • 45
    Di Lelio A, Cestari C, Lomazzi A, Beretta L. Cirrhosis: diagnosis with sonographic study of the liver surface. Radiology 1989; 172: 38992.
  • 46
    Batts KP, Ludwig J. Chronic hepatitis: an update on terminology and reporting. Am J Surg Pathol 1995; 19: 140917.
  • 47
    Lindon JC, Nicholson JK, Everett JR. NMR spectroscopy of biofluids. Ann Rep NMR Spectrosc 1999; 38: 288.
  • 48
    Siegers CP, Watkins JB III. Biliary Excretion of Drugs and Other Chemicals. Stuttgart, Germany: Gustav-Fischer, 1991.
  • 49
    Groen AK, Goldhoorn GG, Egbers PH, Chamuleau RA, Tytgat GN, Bovee WM. Use of 1H NMR to determine the distribution of lecithin between the micellar and vesicular phases in model bile. J Lipid Res 1990; 31: 131521.
  • 50
    Khan SA, Cox IJ, Thillainayagam AV, Bansi DS, Thomas HC, Taylor-Robinson SD. Proton and phosphorus-31 nuclear magnetic resonance spectroscopy of human bile in hepatopancreaticobiliary cancer. Eur J Gastroenterol Hepatol 2005; 17: 7338.
  • 51
    Prescot AP, Collins DJ, Leach MO, Dzik-Jurasz AS. Human gallbladder bile: noninvasive investigation in vivo with single-voxel 1H MR spectroscopy. Radiology 2003; 229: 58792.
  • 52
    Dzik-Jurasz AS, Prescot AP, Leach MO, Collins DJ. Non-invasive study of human gall bladder bile in vivo using (1)H-MR spectroscopy. Br J Radiol 2003; 76: 4836.
  • 53
    Bates TE, Williams SR, Gadian DG. Phosphodiesters in the liver: the effect of field strength on the 31P signal. Magn Reson Med 1989; 12: 14550.
  • 54
    Van Den Bosch H. Phosphoglyceride metabolism. Annu Rev Biochem 1974; 43: 24377.
  • 55
    Maris JM, Evans AE, McLaughlin AC, et al. 31P nuclear magnetic resonance spectroscopic investigation of human neuroblastoma in situ. N Engl J Med 1985; 312: 15005.
  • 56
    Ruiz-Cabello J, Cohen JS. Phospholipid metabolites as indicators of cancer cell function. NMR Biomed 1992; 5: 22633.
  • 57
    Evanochko WT, Sakai TT, Ng TC, et al. NMR study of in vivo RIF-1 tumors. Analysis of perchloric acid extracts and identification of 1H, 31P and 13C resonances. Biochim Biophys Acta 1984; 805: 1046.
  • 58
    Soper R, Himmelreich U, Painter D, et al. Pathology of hepatocellular carcinoma and its precursors using proton magnetic resonance spectroscopy and a statistical classification strategy. Pathology 2002; 34: 41722.
  • 59
    Kuo YT, Li CW, Chen CY, Jao J, Wu D, Liu GC. In vivo proton magnetic resonance spectroscopy of large focal hepatic lesions and metabolite change of hepatocellular carcinoma before and after transcatheter arterial chemoembolization using a 3.0 T MR scanner. J Magn Reson Imaging 2004; 19: 598604.
  • 60
    Dzik-Jurasz AS, Murphy PS, George M, et al. Human rectal adenocarcinoma: demonstration of 1H-MR spectra in vivo at 1.5 T. Magn Reson Med 2002; 47: 80911.
  • 61
    Smith TA, Glaholm J, Leach MO, Machin L, McCready VR. The effect of intra-tumour heterogeneity on the distribution of phosphorus-containing metabolites within human breast tumours: an in vitro study using 31P NMR spectroscopy. NMR Biomed 1991; 4: 2627.
  • 62
    Cadoux-Hudson TA, Blackledge MJ, Rajagopalan B, Taylor DJ, Radda GK. Human primary brain tumour metabolism in vivo: a phosphorus magnetic resonance spectroscopy study. Br J Cancer 1989; 60: 4306.
  • 63
    Smith TA, Glaholm J, Leach MO, et al. A comparison of in vivo and in vitro 31P NMR spectra from human breast tumours: variations in phospholipid metabolism. Br J Cancer 1991; 63: 5146.
  • 64
    Canioni P, Quistorff B. Liver physiology and metabolism. GilliesRJ, ed. In: NMR in Physiology and Biomedicine. New York: Academic Press, 1994; 37388.
  • 65
    Bolan PJ, Meisamy S, Baker EH, et al. In vivo quantification of choline compounds in the breast with 1H MR spectroscopy. Magn Reson Med 2003; 50: 113443.
  • 66
    Griffiths JR, Cady E, Edwards RH, McCready VR, Wilkie DR, Wiltshaw E. 31P-NMR studies of a human tumour in situ. Lancet 1983; I: 14356.