Two-dimensional MR spectroscopy of minimal hepatic encephalopathy and neuropsychological correlates in vivo


  • Aparna Singhal MD,

    1. Department of Radiological Sciences, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    Search for more papers by this author
  • Rajakumar Nagarajan PhD,

    1. Department of Radiological Sciences, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    Search for more papers by this author
  • Charles H. Hinkin MD,

    1. Department of Psychiatry, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    2. VA Greater Los Angeles Healthcare System, Los Angeles, California, USA
    Search for more papers by this author
  • Rajesh Kumar PhD,

    1. Department of Neurobiology, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    Search for more papers by this author
  • James Sayre PhD,

    1. Department of Radiological Sciences, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    2. Department of Public Health-Biostatistics David Geffen School of Medicine, University of California, Los Angeles, California, USA
    Search for more papers by this author
  • Virginia Elderkin-Thompson MD,

    1. Department of Psychiatry, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    Search for more papers by this author
  • Amir Huda PhD,

    1. Department of Radiological Sciences, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    2. Department of Physics, California State University, Fresno, California, USA
    Search for more papers by this author
  • Rakesh K. Gupta MD,

    1. Department of Radiology, Sanjay Gandhi Post-Graduate Institute of Medical Sciences, Lucknow, India
    Search for more papers by this author
  • Steven-Huy Han MD,

    1. Department of Hepatology, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    Search for more papers by this author
  • M. Albert Thomas PhD

    Corresponding author
    1. Department of Radiological Sciences, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    2. Department of Psychiatry, David Geffen School of Medicine, University of California, Los Angeles, California, USA
    • Department of Radiological Sciences, David Geffen School of Medicine at UCLA, CHS BL 428, 10833 Le Conte Avenue, Los Angeles, CA 90095-1721
    Search for more papers by this author

  • Presented at the 17th International Society of Magnetic Resonance in Medicine (ISMRM) Meeting, Honolulu, Hawaii, April 18–25, 2009.



To evaluate regional cerebral metabolic and structural changes in patients with minimal hepatic encephalopathy (MHE) using two-dimensional (2D) MR spectroscopy (MRS) and Tmath image-weighted MRI, to correlate the observed MR changes with neuropsychological (NP) test scores, and to compare the diagnostic accuracy of MRI, 2D MRS, and NP tests in discriminating between patients and healthy subjects.

Materials and Methods:

Thirty-three MHE patients and 30 healthy controls were investigated. The 2D localized correlated spectroscopy (L-COSY) was performed in the frontal and occipital brain on a 1.5 Tesla (T) MR scanner. The NP test battery included 15 tests, grouped into 6 cognitive domains. Globus pallidus signal intensities were calculated from T1-weighted images.


The 2D MRS showed significant differences in ratios of the following metabolite(s) peaks with respect to creatine (Cr): decreased myo-inositol (mI), choline (Ch), mICh, and increased (glutamate plus glutamine) (Glx) in patients compared with healthy subjects in both occipital and frontal lobes. Frontal lobe taurine also showed a decline in patients. The NP test results revealed declines in cognitive speed, motor function, executive function, and global cognitive status. Significant correlations were found between the altered metabolites and NP tests. Alteration in the mICh/Cr ratio was noted as a powerful discriminant between healthy subjects and the patients.


The study demonstrates that relative metabolite levels determined by 2D MRS, in particular mICh/Cr, provide the best diagnostic prediction for MHE. The results suggest that depletions of myo-inositol, choline and taurine with respect to creatine correlate with measures of neuropsychological impairment. J. Magn. Reson. Imaging 2010;32:35–43. © 2010 Wiley-Liss, Inc.

HYPERAMMONEMIA HAS CONSISTENTLY been shown to be one of the major contributory factors in the pathogenesis of hepatic encephalopathy (HE). Current hypotheses support the ammonia-glutamine-low grade cerebral edema hypothesis of neuropsychological dysfunction in cirrhosis (1). Astrocytes are the sites of ammonia detoxification in the brain through the glutamate-glutamine cycle. 1H magnetic resonance spectroscopy (MRS) enables the recording of the levels of glutamate plus glutamine (Glx) in vivo (2, 3). In patients with HE, the observation that myo-inositol (mI) is reduced with increasing concentrations of Glx suggests that mI may be an important osmotic regulator within the astrocyte (4, 5). Recent studies in patients with cirrhosis have shown that the metabolic disturbances are associated with increased brain water, the severity of which correlates with worsening of neuropsychological function (1, 6–8). Apart from the reported metabolic changes, T1-weighted MRI shows globus pallidal (GP) hyperintensities (9, 10).

Patients with cirrhosis and HE frequently demonstrate various neuropsychiatric symptoms and deficits. The pathophysiology, natural history, and prognosis of cirrhosis-associated neuropsychiatric deficits are not completely established, and the data on this issue are controversial (11, 12). Recent studies have shown that these deficits are associated with changes in metabolic brain patterns and may reverse with the reduction in blood ammonia level (13). On the other hand, long-term persistence of these symptoms after liver transplantation has been reported (14, 15), and a neurodegenerative nature of this disorder has been suggested (15). In addition, cirrhosis-associated neurological symptoms, patients with hepatic encephalopathy frequently demonstrate bioregulative (disturbed sleep and sexual dysfunction), behavioral, and affective symptoms (11, 16, 17). The adequate clinical neuropsychiatric evaluation of these patients remains difficult because, in part, neuropsychiatric symptoms associated with low grade encephalopathy are multiform and sometimes subtle. This problem is further accentuated in the case of minimal HE (MHE) which is diagnosed in patients who demonstrate deficits only on neuropsychological tests but are normal upon clinical neuropsychiatric examination (11, 16, 17). Thus, evaluating patients with MHE remains a challenge.

One-dimensional (1D) and 2D MRS in vivo derived metabolites ratios of mI and Glx with respect to creatine have been shown to correlate with the grade of encephalopathy as well as neuropsychological (NP) tests in low grade HE (3, 7, 18). However, most 1D MRS studies in the literature have used the single voxel approach with two voxels placed on the parietal lobe and basal ganglia. On studying the anterior cingulate gyrus in minimal HE using 2D MRS, Binesh et al found significant decrease in mI and choline (Ch) and increase in Glx (18). In the present study, we performed 2D localized correlated spectroscopy (L-COSY) acquired over two voxels (right frontal lobe and left occipital lobe), with an aim to correlate the relative metabolite concentrations with the region specific NP tests. Another goal was to compare the diagnostic accuracy of 2D MRS, MRI signal intensities and NP tests, alone and in combination, in discriminating between minimal HE patients and controls. To the best of our knowledge, this is the first study that correlates the regional changes in the metabolites using 2D MRS with NP tests in patients with minimal HE.


Human Subjects

We studied 33 patients (15 females and 18 males) with a mean age of 50.5 ± 9.4 years diagnosed with minimal hepatic encephalopathy and 30 healthy controls (17 females and 13 males) with a mean age of 48.9 ± 11.8 years. We included patients listed for orthotopic liver transplantation (OLT) and with a United Network Organ Sharing (UNOS) status of 3 or greater. The patients were clinically evaluated at the time of initial assessment. Initial cognitive evaluation included a psychiatric history and the Mini-Mental Status Examination (MMSE). Patients scoring < 24 on the MMSE were classified as cognitively impaired and were excluded from further study (19). Participants were administered a battery of neuropsychological tests selected for their presumed sensitivity to the neurocognitive effects of HE, the details of which are described later. Patients with no evidence of manifest HE, according to their mental state (West-Haven criteria) (11), were classified as having minimal HE when 2 or more of the tests were abnormal (12). The etiologies of the patients' liver disease were: alcoholic cirrhosis (N = 7), primary biliary cirrhosis (N = 2), autoimmune hepatitis (N = 1), Budd Chiari syndrome (N = 1), primary sclerosing cholangitis (N = 1), and chronic hepatitis C cirrhosis (N = 21), including a history of concomitant alcoholic liver disease in 1 patient. Exclusion criteria of the patients were ages less than 18 or more than 75 years, active alcoholism during the 3 months before the study, HE grades II–IV, additional neurological or psychiatric disease, treatment with psychotropic drugs or other drugs known to alter neuropsychological function, gastrointestinal bleeding or infection within 1 week before the study, previous treatment with shunt procedures. Control subjects were recruited through advertisements. None of the controls had any history of neurological or psychiatric illness, metabolic disorders, alcohol or drug abuse, head injury, or liver disease. All underwent detailed clinical assessment, including neurological examination. Control and HE subjects with contraindication for MRI scanner environment were also excluded from the study. The exclusion criteria also included a lack of fluency in the English language (affecting psychometric tests). The Institutional Review Board (IRB) approved the protocol, and written informed consent was obtained from each subject.

Neuropsychological Tests

Neuropsychological testing was conducted on the same day as the MRI/MRS scan and took approximately 2 hours to complete. The tests were grouped into domains according to the ability they tested. The assessed domains with the respective tests in the NP battery included: (i) Premorbid Estimate of Verbal Intelligence: Wechsler test of adult reading; (ii) Learning and Memory: revised Hopkins verbal learning test and revised brief visuospatial memory test; (iii) Cognitive Speed/Speed of Information Processing: WAIS-III Digit Symbol, WAIS-III Symbol Search and Trail Making test Part A; (iv) Abstraction/Executive Functioning: Wisconsin Card Sorting test (64-item version), Trail Making test Part B and Stroop Color Word test; (v) Attention/Working Memory: Continuous Performance test-II, Paced Auditory Serial Addition Task-50 and WAIS-III Letter-Number Sequencing; (vi) Visuospatial Function: WAIS-III Block Design; (vii) Motor: Grooved Pegboard (Dominant and Nondominant); (viii) Language: Letter Fluency test (FAS); (ix) Psychiatric Status: Brief Symptom Inventory and Beck Depression Inventory-II (19). All raw test scores were then converted to demographically-corrected T scores using extant normative data (20). Domain T scores were obtained by calculating the mean T score for all the tests that comprised that neurocognitive domain. A Global T score was then calculated by taking the average of the individual test scores. Subjects were classified as “impaired” on a particular cognitive domain if the domain T score was less than 40 (i.e., at least 1 standard deviation in the impaired direction).

MRI/MRS Acquisition and Postprocessing

We used a 1.5 Tesla (T) Avanto MR scanner (Siemens Medical Solutions, Erlangen, Germany). The subjects were imaged while lying supine using a quadrature body coil for “transmission” and a dual surface coil used for “reception” for 2D MRS. One surface coil was placed on the right forehead and the second coil on the left occipital region of the subject. Using a dual-echo turbo spin-echo pulse sequence, proton density (PD) and T2-weighted images were acquired in the axial plane covering the entire brain using a 12-channel “receive” head coil and the following parameters: repetition time (TR) = 7500 ms; echo-times (TE1, TE2) = 17 and 133 ms; matrix size = 256 × 256; field of view = 230 × 230 mm2; slice thickness = 4.0 mm; interslice gap = 0.0 mm; turbo factor = 5; and flip angle = 150°. High-resolution T1-weighted images were collected using a magnetization prepared rapid acquisition gradient-echo (MPRAGE) pulse sequence (TR = 1660 ms; TE = 3.87 ms; inversion time = 900 ms; flip angle = 10°; matrix size = 256 × 256; field of view = 230 × 230 mm2; slice thickness = 1.2 mm; number of slices= 176). MR Spectroscopy was performed using the 2D L-COSY sequence(21) and the experimental parameters were as follows: voxel size of 27 cm3, TR/TE = 2 s/30 ms, 1024 complex points and spectral width of 2000 Hz along the t2 dimension, 96 Δt1 increments and 625 Hz along the indirect t1 dimension and 8 averages per Δt1. A WET-based sequence was used for global water suppression (22). The voxel shim was done manually, and the range of water resonance line width was 8–10 Hz. The total duration for each two-dimensional scan was approximately 25 min. The 2D MRS data were processed using the Felix NMR software (Felix NMR, CA, USA) that is more commonly used in high resolution NMR. The raw MRS data matrix was apodized with squared sinebell filters along the two axes, and the entire data matrix was zero-filled to 2048 × 256. A double Fourier transformation resulted in a 2D L-COSY spectrum. The 2D spectral peaks were displayed as contour plots in the magnitude mode. Raw volume integrals of the 2D spectral peaks were measured and using these volumes, ratios of the diagonal (_d) and cross peaks of different metabolites were calculated with respect to the diagonal peak of creatine (Cr_d) (F2 = F1 = 3.0 ppm). The mean and standard deviation of the integral limits for the volume calculation of diagonal peaks (NAA_d, Cr_d, Ch_d, and mI_d) and cross peaks (Glx, Asp, NAA, PE, PCh, Tau, ThrLac, GABA, mICh) were 0.168 ± 0.047 ppm and 0.239 ± 0.197 ppm.


MRI data were processed using the statistical parametric mapping package (SPM2; Wellcome Department of Cognitive Neurology, UK), and a Matlab-based (MathWorks Inc, Natick, MA) custom software. High-resolution T1-weighted images were normalized to the standard Montreal Neurological Institute (MNI) T1-weighted template. Bilateral whole GP masks were created manually using normalized T1-weighted images from a healthy subject. Median signal intensity from T1-weighted images was calculated in each subject using these masks, and relative change in T1 signal intensity in the patient group compared with controls was calculated.

The 2D MRS evaluation was performed in the following locations: (i) prefrontal dorsolateral white/gray matter and (ii) occipito-parietal white/gray matter. Our earlier work on 2D MRS of HE had sampled the medial frontal gray matter using the GE 1.5T MRI scanner. Because we expect to see global changes of cerebral metabolites in HE, we chose the frontal white and occipito-parietal regions in this project using the Siemens 1.5T MRI scanner.

The first location was investigated in 24 patients and 18 healthy controls and the 2nd location in 20 patients and 16 healthy controls. Of all the subjects, 19 patients and 20 healthy controls were evaluated by MRI, and 25 patients and 27 controls underwent neuropsychological testing. Of these, 14 patients and 14 controls each underwent neuropsychological tests as well as MRS evaluation in both frontal and occipital regions.

Statistical Analysis

Data were analyzed using SPSS 15.0 for Windows (SPSS Inc., Chicago, IL). Results were expressed as (mean ± SD). Differences between patients and healthy controls were tested using the two-tailed t-test. Within each group, differences were tested with the paired t-test. A P-value < = 0.05 was considered statistically significant and a Bonferroni adjustment was used. To explore relationships between MRS ratios, MRI findings, and NP T-scores, Pearson correlation was performed. A multiple regression analysis was also performed to assess correlation between MRS and NP domains to account for shared covariance. The covariates included age, gender, all the metabolite ratios and NP T-scores. A stepwise logistic regression analysis was performed on the metabolites ratios to test if a combination of these ratios could possibly separate healthy controls from patients. Variables were selected with the same level of significance (P ≤ 0.05). The performance of the algorithm was first evaluated by a receiver operating characteristic (ROC) analysis (23), allowing the cut point value to be changed over a range in a set of repeated experiments. The resulting pairs of sensitivity and specificity for each cut point were examined and plotted as a set of points connected by a curve. The area under the ROC curve (AUROC) reflects the performance of the method used to classify the lesions. A specific cut point value had to be defined to discriminate controls from patients with a reasonable sensitivity to calculate the specificity, positive predictive value, and accuracy of the classification scheme.


2D MR Spectroscopy

Figure 1 shows a 2D L-COSY spectrum recorded in the occipital lobe of a 51-year-old patient. Table 1 summarizes the mean ratios (± SD) of the metabolites with respect to creatine in MHE patients and healthy controls. Ratios of frontal and occipital lobe mICh (overlapping cross peak of myo-inositol and choline), mI, Ch_d, and frontal lobe Tau were significantly lower in patients compared with healthy controls, whereas frontal and occipital lobe Glx ratios were significantly higher (P < 0.05). Occipital lobe Tau ratio showed only a trend toward decline (P > 0.05). The distributions of frontal and occipital lobe mICh/Cr_d are shown in Figure 2.

Figure 1.

The 2D L-COSY spectrum from the occipital lobe of a 51-year-old patient. Cr, creatine; Ch, choline; Glx, glutamate plus glutamine; NAA, N-acetyl aspartate; Asp, aspartate; PE, phosphoethanolamine; PCh, phosphocholine; mI, myoinositol; Tau, taurine; ThrLac, overlapping cross peaks of threonine and lactate; GABA, gamma-aminobutyric acid; mICh, overlapping cross peaks of myoinositol and choline; MM, macromolecules. Diagonal peaks of the respective metabolites are labeled as _d.

Table 1. Metabolite Ratios (Mean ± SD) Calculated From the 2D MRS Peak Volumes
 Ratios to Cr_dControlsPatientsP value
  • *

    P value < 0.05.

  • Cr_d = diagonal creatine peak; Ch_d = diagonal choline peak; NAA_d = diagonal N-acetyl aspartate peak; mICh = overlapping cross peak of myo-inositol and choline; Asp = aspartate; Glx = glutamate plus glutamine; mI = myo-inositol; Tau = taurine; PCh = phosphocholine; GABA = gamma-aminobutyric acid; PE = phosphoethanolamine; ThrLac = overlapping cross peaks of threonine and lactate.

Frontal lobeCh_d0.992 ± 0.1080.867 ± 0.1580.006*
NAA_d1.618 ± 0.2181.54 ± 0.2090.263
mICh0.115 ± 0.0120.069 ± 0.0240.000*
Asp0.026 ± 0.0050.024 ± 0.0060.426
Cr0.015 ± 0.0040.014 ± 0.0050.223
Glx0.118 ± 0.0160.153 ± 0.0260.000*
mI0.023 ± 0.0050.013 ± 0.0060.000*
NAA0.319 ± 0.0450.308 ± 0.0610.537
Tau0.010 ± 0.0030.008 ± 0.0020.011*
PCh0.007 ± 0.0020.006 ± 0.0030.815
GABA0.007 ± 0.0030.008 ± 0.0040.328
PE0.007 ± 0.0020.006 ± 0.0030.306
ThrLac0.011 ± 0.0020.013 ± 0.0040.148
Occipital lobeCh_d0.934 ± 0.1070.773 ± 0.1640.002*
NAA_d1.62 ± 0.1161.532 ± 0.1820.103
mICh0.115 ± 0.0180.066 ± 0.0260.000*
Asp0.026 ± 0.0050.027 ± 0.0080.404
Cr0.014 ± 0.0040.016 ± 0.0050.209
Glx0.116 ± 0.0150.159 ± 0.0510.003*
mI0.021 ± 0.0050.015 ± 0.0070.007*
NAA0.299 ± 0.0430.282 ± 0.0670.379
Tau0.009 ± 0.0020.008 ± 0.0030.337
PCh0.007 ± 0.0020.007 ± 0.0030.612
GABA0.007 ± 0.0020.007 ± 0.0050.907
PE0.004 ± 0.0010.006 ± 0.0040.115
ThrLac0.014 ± 0.0070.012 ± 0.0070.508
Figure 2.

Distribution of the mICh/Cr_d ratio in patients and healthy controls in (a) frontal and (b) occipital regions.


Mean (±SD) GP signal intensity measured from T1-weighted images in patients [left GP: 410.47 (±48.91), right GP: 413.60 (±49.09)] was significantly higher than that of healthy controls [left GP, 350.52 (±28.51); right GP, 350.07 (±25.69)] (P < 0.0001).

Neuropsychological Assessment

Performance on the NP tests is summarized in Figure 3. Patients showed significantly lower scores than the healthy controls for the following domains: cognitive speed, motor function, executive function, and global scores (P < 0.05). Amongst all domain scores in patients, motor (34.38 ± 12.20) and cognitive speed (35.40 ± 7.88) scores were the lowest.

Figure 3.

The mean T-scores of the individual neuropsychological domains. The standard deviation is shown by the error bars.

Table 2 summarizes the correlation coefficients between the NP domain scores and MRS ratios in the patients (n = 14). Frontal and occipital lobe mICh and Tau and frontal lobe mI ratios positively correlated with motor and cognitive speed domain scores. Frontal lobe Ch_d ratio positively correlated with the cognitive speed domain scores. Glx ratio negatively correlated with the significantly altered executive domain. On performing multiple regression analysis, frontal lobe mI, Ch_d, mICh, and taurine and occipital lobe mICh ratios positively correlated with cognitive speed, whereas Glx negatively correlated with the language domain. No significant correlation was found between the GP signal intensity and the NP test results or the 2D MRS metabolite measures.

Table 2. Statistical Correlation of 2D MRS Ratios With the Neuropsychological Domain Scores Using the Regression Analysis
RatiosNP domainCorrelation coefficient
  • *

    Significant at 0.05.

  • **

    Significant at 0.01.

  • ***

    Only multiple regression to predict Ch_d/Cr was 0.829 (CogSpeed + Motor).

  • NP = Neuropsychological; CogSpeed = Cognitive Speed; Cr = creatine; mI = myo-inositol; Glx = glutamate plus glutamine; Ch_d = diagonal choline peak; mICh = overlapping cross peak of myo-inositol and choline; Tau = taurine; Asp = aspartate; NAA_d = diagonal N-acetyl aspartate peak; ThrLac = overlapping cross peaks of threonine and lactate; PCh = phosphocholine; GABA = gamma-aminobutyric acid.


A logistic regression analysis was performed on all the modalities and their combinations (Fig. 4). Among the frontal lobe MRS ratios, the two ratios that gave the best predictability were mICh and Glx, in order of importance with their respective cutoff values as 0.10 and 0.12. mICh was the only significant ratio in the occipital region, with a cutoff value of 0.11. Cognitive speed was the selected domain among the NP tests with a cutoff value of 40. The respective sensitivities, specificities, negative and positive predictive values are given in Table 3. On combining frontal and occipital lobe MRS, the selected variables were frontal and occipital lobe mICh and had a predictability of 100% in this pilot study with limited number of subjects. Performance of the logistic regression results were analyzed by constructing ROC curves and comparing the AUROCs (Fig. 5). The AUROC of the best model (frontal + occipital mICh) was 1. Next best was frontal lobe mICh alone at 0.98.

Figure 4.

Diagnostic accuracies of different modalities and their combinations in correctly classifying the overall population and patients and controls separately, based on logistic regression analysis. The x-axis values represent a modality or a combination followed by the significant variable(s) in parentheses. *For globus pallidus signal intensities, neither side was selected as a significant variable, Cog Speed, cognitive speed.

Table 3. Specificity, Sensitivity, and Predictive Values of the Significant Variables
 Frontal MRSOccipital MRSNP tests
  1. PPV = positive predictive value; NPV = negative predictive value; Cog Speed = cognitive speed.

Significant variable(s)mICh and GlxmIChCogSpeed
Figure 5.

Results of logistic regression analyses and ROC curves predicting the sensitivity and specificity using the following tests: (a) neuropsychological domain scores; (b) frontal lobe MRS; (c) occipital lobe MRS; (d) neuropsychological domain scores combined with the occipital lobe metabolite ratios; (e) neuropsychological domain scores combined with the frontal lobe metabolite ratios; adding occipital lobe MRS ratios to this did not change the results. Combining both frontal and occipital lobe MRS ratios correctly predicted 100% of the subjects by selecting occipital lobe mICh/Cr_d and frontal lobe mICh/Cr_d as the significant variables (AUROC Curve = 1, curve not shown). For all curves, the asymptotic significance level was less than 0.0001.

Regional MRS Differences

There were significant differences in some of the MRS ratios between the two locations within the same subject (P < 0.05). In healthy volunteers, mean Ch_d and PE ratios were lower in the occipital lobe (Ch_d: 0.934 ± 0.107; PE: 0.0044 ± 0.0012) than the frontal lobe (Ch_d: 1.012 ± 0.096; PE: 0.007 ± 0.002) and in the patient population, mean occipital lobe NAA and Ch_d ratios (NAA: 0.282 ± 0.067; Ch_d: 0.773 ± 0.164) were lower as compared to frontal region ratios (NAA: 0.316 ± 0.064; Ch_d: 0.867 ± 0.166). In patients, occipital lobe Cr (0.027 ± 0.008) had higher mean ratios than that in the frontal region (0.014 ± 0.005).


The current study attempts to correlate different modalities, namely 2D MRS, MRI, and NP tests, in minimal HE and to evaluate their efficacy to accurately detect patients of minimal HE. Our results demonstrate that 2D MRS ratios provide the maximum diagnostic accuracy in differentiating MHE patients from healthy controls as compared to either NP tests or MRI alone, with one ratio (mICh) being the most discriminant variable. A combination of frontal and occipital lobe mICh gave the highest diagnostic accuracy (100%). Comparing them independently, frontal lobe MRS ratios had higher predictability than occipital ratios.

Early diagnosis of MHE might be of particular importance considering the fact that after a follow-up period of 3 years, around 50% of cirrhotic patients with minimal HE present with clinically overt HE as opposed to 8% of the patients without MHE (24). This underscores the prognostic significance of MHE, in addition to its adverse impact on the quality of life of the patients. It has been shown that MHE patients show psychomotor slowing and executive dysfunction, whereas other cognitive abilities are relatively preserved, reflecting a disorder mainly affecting the prefrontal cortex and its connections with the basal ganglia (17, 25, 26). Our results also showed impairments in cognitive speed, motor function, and executive function leading to an overall decline in global cognitive domain score in the MHE patients.

According to the most plausible hypothesis for pathogenesis of HE, astrocytes convert glutamate to glutamine to detoxify increased ammonia in cirrhotic patients. The increased glutamine causes an osmotic imbalance leading to cellular swelling and cerebral edema which in turn is related to the neuropsychological manifestations. This cellular volume disturbance leads to compensatory release of osmolytes such as myo-inositol (4, 5, 27). In agreement with the previous studies (18, 28–31), increased Glx and decreased mI observed in frontal and occipital regions of MHE patients, are consistent with this current hypothesis. The reduction of choline ratio in HE, however, is not very well understood. The 2D L-COSY is a technique that allows less ambiguous assignment of the metabolites already observed on 1D MRS, such as Glx, and, furthermore, can resolve additional J-coupled metabolites like Asp, Tau, and GABA. At clinical field strengths, it can partially resolve various subgroups of choline (18). We did not find any significant differences in PCh or PE ratios between healthy subjects and patients. The choline in the mICh cross peak which showed a significant decrease in MHE patients, represents free choline. Free choline is made available from the membrane phospholipid, phosphatidylcholine, and is transported from the circulation (32). Although changes in PCh and PE might not be detected sensitively using 2D MRS at 1.5T because of their lower concentrations compared with other metabolites, the current data suggests that the absence of significant changes in PCh and PE ratios could signify an unaltered phospholipid metabolism. This could suggest that the decrease in free choline results from either its function as an osmolyte (33, 34) or a decreased supply of choline due to its decreased de novo synthesis in the liver or a decreased uptake into the brain due to altered transport (35, 36).

In the current study, we also observed a significant decline in the frontal lobe taurine ratio of patients. Taurine is an abundant amino acid present in the central nervous system (CNS) and thought to be a weak GABA agonist. It has also been suggested to be an osmoregulator, a neuroprotectant, and an antioxidant (37, 38). It had not been possible to study in vivo human brain taurine until recently because of the lack of noninvasive means to do so. In humans, taurine had earlier been demonstrated in autopsied brain tissue of patients dying with hepatic coma (39). Brain taurine content is decreased in rats with portocaval anastamosis serving as a model for chronic HE (6) and in cultured astrocytes but not in neurons (40). In several experimental studies, decreased taurine content has been shown to correlate with cerebral edema (6, 37, 41). Thus, evidence of in vivo decrease in taurine lends further support to the ammonia-glutamine-low grade cerebral edema hypothesis as this decrease in taurine possibly occurs due to a compensatory loss to maintain cell volume homeostasis. However, ammonia induced taurine release has also been seen due to its activation of ionotropic glutamate receptors unrelated to cell volume regulatory changes (42).

Several significant correlations were observed between the metabolite ratios and the NP domain scores. These strong and significant correlations between the altered metabolites and the implicated NP domains could mean that the metabolic alterations have functional consequences and they also lead to neuropsychological impairment. It is possible that the clinical manifestations in MHE are mediated by specific neurochemical alterations. Motor function scores positively correlated with frontal lobe mI, mICh, and Tau and occipital lobe mICh ratios. Frontal lobe choline ratio correlated only with the cognitive speed domain and not the motor, and frontal lobe mI ratio showed a slightly stronger correlation than the mICh ratio to both the domains. This suggests that changes in myo-inositol may represent metabolic changes different than those represented by choline and that both may have different functional impacts. Frontal lobe Tau was found to correlate with the significantly altered domains of motor function and cognitive speed, thus strengthening in vivo evidence for its role in the pathophysiology of HE. Taurine transport can modulate neuronal excitability, synaptic transmission, and plasticity, and taurine has also been suggested to play a role in organization of movement (43–45). In in vivo models of HE, Tau release and redistribution has been suggested to promote inhibitory neurotransmission and/or cell membrane protection (37). Future in vivo studies using 2D MRS with different grades of HE might be able to further evaluate the role of taurine in the pathogenesis and progression of HE.

GP hyperintensities have been consistently reported on T1 weighted images in HE (9), and they have been postulated to be due to the deposition of manganese (46, 47). In majority of studies done earlier, GP hyperintensities have been shown to have no correlation with the neuropsychiatric impairment (10, 13, 48), whereas a correlation has been seen with plasma ammonia and blood manganese (49–51). There has been no consensus about the correlation of hyperintensities with biochemical findings (32, 52). The differences between various studies may have arisen due to heterogeneous study groups, differences in MRS methodology, studied brain locations, and methods of evaluating the T1 weighted images. In our study also, there was no correlation between GP signal intensity and 2D MRS ratios or NP test scores, although the signal intensities were significantly higher in the patients compared with controls. Separation of patients from controls using MRI changes on both sides was not possible using the logistic regression analysis. This possibly suggests that this MRI finding is a concurrent change with the neuropsychological and metabolic alterations rather than a causative one and has limited diagnostic significance.

Compared with our earlier report where the medial frontal gray/white matter was investigated (18), the trend of metabolite ratio changes of Ch_d, mI, Glx, NAA_d, mICh, Cr and GABA was similar in the current work investigating the frontal and the occipital lobe white/gray matter regions. However, other metabolite ratios except Tau showed nonsignificant opposite trends. In the frontal lobe, the ratio of Tau was significantly decreased in HE patients compared with healthy controls in agreement with a previous report on autopsied brain tissue from patients with chronic liver disease (53). It showed similar trend in the occipital lobe without any significant change. Even though the experimental parameters were more or less the same, 1.5T MRI scanners manufactured by two different vendors were used in these investigations.

With the feasibility of in vivo 2D MRS in HE, more metabolic markers may be assessed in larger studies with the use of higher fields. Perhaps, a more reliable quantification procedure such as prior knowledge fitting algorithm (ProFit) to process 2D spectra might help in more accurate evaluation of low concentrations of J coupled metabolites (54). Future studies might elucidate the underlying role of various metabolites in different neurocognitive functions, and evaluation of in vivo metabolic markers can also facilitate identification of therapeutic targets in the disease.

It is worth mentioning that our study had some limitations. First, the study was performed in only one grade of HE. Thus, further extension of the technique in a larger group of patients with varying grades of HE is essential. Second, although the cross peak of Glx was much better separated from other overlapping metabolites in 2D L-COSY compared with 1D MRS studies, glutamate could not be separated from glutamine because of the almost identical peak locations. Our recent progress on differentiating Glu from Gln using the ProFit quantitation of 2D L-COSY spectra will be reported elsewhere. Third, it was assumed that Cr levels did not change in HE compared with healthy subjects. In localized proton MRS, metabolites levels have often been expressed as ratios, rather than as absolute concentrations. Fourth, the relationship between T1/T2 and the observed metabolic ratio changes needs to be further explored. The lack of correlation between MRI findings with other variables could be due to a limited number of patients available for this statistical analysis and might need to be probed in a larger cohort of subjects.

In conclusion, our results demonstrate that the 2D MRS ratios not only provide pathophysiological information and correlate with neuropsychological impairment but also give the best diagnostic predictability for patients with MHE. The ratio of mICh to creatine stands out as a powerful discriminant, being both more sensitive and specific than NP tests or MRI in predicting HE. With the application of 2D MRS techniques, unlike 1D MRS, the J-coupled metabolites such as taurine can be resolved and evaluated in vivo. At present, the availability of an MRI scanner that enables the 2D MRS technique might be a limiting factor in its wide applicability as compared to neuropsychological testing, but considering its high diagnostic predictability in patients with MHE, potential avenues for future research and clinical applications should open.


Scientific support of Drs. Nader Binesh, Shida Banakar, and Sherry Liu during the earlier phase of this project is gratefully acknowledged.