Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Is MR Spectroscopy Really the Best MR-Based Method for the Evaluation of Fatty Liver in Diabetic Patients in Clinical Practice?

Is MR Spectroscopy Really the Best MR-Based Method for the Evaluation of Fatty Liver in Diabetic Patients in Clinical Practice?

  • Daniella Braz Parente, 
  • Rosana Souza Rodrigues, 
  • Fernando Fernandes Paiva, 
  • Jaime Araújo Oliveira Neto, 
  • Lilian Machado-Silva, 
  • Valeria Lanzoni, 
  • Carlos Frederico Ferreira Campos, 
  • Antonio Luis Eiras-Araujo, 
  • Pedro Emmanuel Alvarenga Americano do Brasil, 
  • Philippe Garteiser



To investigate if magnetic resonance spectroscopy (MRS) is the best Magnetic Resonance (MR)-based method when compared to gradient-echo magnetic resonance imaging (MRI) for the detection and quantification of liver steatosis in diabetic patients in the clinical practice using liver biopsy as the reference standard, and to assess the influence of steatohepatitis and fibrosis on liver fat quantification.


Institutional approval and patient consent were obtained for this prospective study. Seventy-three patients with type 2 diabetes (60 women and 13 men; mean age, 54±9 years) underwent MRI and MRS at 3.0 T. The liver fat fraction was calculated from triple- and multi-echo gradient-echo sequences, and MRS data. Liver specimens were obtained in all patients. The accuracy for liver fat detection was estimated by receiver operator characteristic (ROC) analysis, and the correlation between fat quantification by imaging and histolopathology was analyzed by Spearman's correlation coefficients.


The prevalence of hepatic steatosis was 92%. All gradient-echo MRI and MRS findings strongly correlated with biopsy findings (triple-echo, rho = 0.819; multi-echo, rho = 0.773; MRS, rho = 0.767). Areas under the ROC curves to detect mild, moderate, and severe steatosis were: triple-echo sequences, 0.961, 0.975, and 0.962; multi-echo sequences, 0.878, 0.979, and 0.961; and MRS, 0.981, 0.980, and 0.954. The thresholds for mild, moderate, and severe steatosis were: triple-echo sequences, 4.09, 9.34, and 12.34, multi-echo sequences, 7.53, 11.75, and 15.08, and MRS, 1.71, 11.69, and 14.91. Quantification was not significantly influenced by steatohepatitis or fibrosis.


Liver fat quantification by MR methods strongly correlates with histopathology. Due to the wide availability and easier post-processing, gradient-echo sequences may represent the best imaging method for the detection and quantification of liver fat fraction in diabetic patients in the clinical practice.


Non-alcoholic fatty liver disease (NAFLD) affects 10–30% of the general population across all ethnicities and age groups [1][5]. Obesity and diabetes are the primary risk factors. Worldwide, the prevalence of obesity, diabetes, and fatty liver is rising [1], [3][5]. NAFLD is a clinicopathologic syndrome with a wide spectrum of histological abnormalities and clinical outcomes [1], [2], [4], . While some patients have isolated steatosis, others have non-alcoholic steatohepatitis (NASH), which can progress to cirrhosis, and are at increased risk for hepatocellular carcinoma [1], [2], [4], [6], [7]. Today, NAFLD is the third most common indication for liver transplantation in the United States and is on a trajectory to become the most common indication [8]. Therefore, early diagnosis is important for appropriate treatments and to prevent progression.

The prevalence of NAFLD in diabetic patients is approximately 70% [9][12]. Type 2 diabetes is an important risk factor for NAFLD and the disease follows a more aggressive course in these patients with necroinflammation and fibrosis [13][17]. The prevalence of NASH in diabetic patients is not well established [18], but has been estimated to be between 22% and 88% [11], [19]. Diabetic patients commonly have comorbidities (e.g., obesity or coronary heart disease) and may be at higher risk of complications during liver biopsy [20]. Therefore, noninvasive methods to diagnose NAFLD, quantify liver fat, stage disease severity, and monitor patients over time is important.

Several imaging methods have been used in the diagnosis of NAFLD. Ultrasonography (US) is the most common and accurately detects moderate and severe NAFLD, but it is not sensitive to mild steatosis [21][23]. Interobserver agreement for the severity of NAFLD can be as low as 55% [24]. Therefore, US does not allow precise fat liver quantification for patient follow-up. Computed tomography (CT) is also limited for the detection of mild steatosis and ionizing radiation prevents its use for long-term monitoring [21][23], [25]. Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) are considered excellent imaging methods for the noninvasive detection and quantification of liver fat [23], [26][28]. However, the accuracy of MRI and MRS in the detection and quantification of liver fat fraction, the thresholds for mild, moderate and severe steatosis, and the influence of steatohepatitis and fibrosis on the quantification of liver fat fraction are still unknown. Therefore, the goal of this study was to assess the diagnostic performance of triple- and multi-echo MRI, and 1H spectroscopy in the evaluation of hepatic steatosis, to determine the thresholds for different steatosis grades, and to assess if steatohepatitis and/or fibrosis changes MR quantification of liver steatosis using histologic assessment as the reference standard.

Materials and Methods


This prospective study was approved by the Institutional Ethics Committee of Pedro Ernesto University Hospital. Written informed consent was obtained from all patients. Between June 2010 and February 2012, type 2 diabetic patients at the University of the State of Rio de Janeiro, Brazil, between 18 and 70 years of age with clinical indications for liver biopsy to evaluate NAFLD were consecutively enrolled. In order to study a homogeneous population with NAFLD, patients with other chronic liver diseases were excluded, because some of them, like hepatitis C can be associated with fatty liver and with fibrosis. Exclusion criteria were: other possible causes of chronic liver disease (positive serology for hepatitis B or C), history of alcoholism (≥20 g of alcohol per day), severe or decompensated cardiopulmonary disease, renal failure (creatinine >1.5 mg/dL), coagulopathy disorders, medication that could cause NAFLD, contraindications to MRI (i.e., claustrophobia or metallic implants), or refusal of liver biopsy. Criteria for withdrawal from the study included other etiology for chronic liver disease upon liver biopsy or insufficient biopsy material for histological analysis.


MRI and MRS were performed during the same examination on a 3.0 T magnet (Philips Medical Systems, Eindhoven, Netherlands) with a Quasar dual gradient system with a peak gradient amplitude of 80 mT/m and slew rate of 200 mT/m/ms. A whole-body transmitter coil and an sixteen-element, receive-only, phased-array coil were used. All patients were supine. Patients suspended respiration at the end of inspiration for breath-hold sequences and were instructed to breathe smoothly for the respiratory-triggered sequences. For correct positioning, coronal localizer images were acquired at maximum inspiration and expiration. Axial and coronal T2 images were obtained for anatomical reference.

Gradient echo sequences

The liver fat fraction was quantified using T1-weighted, 2D spoiled gradient-echo sequences. Low flip angles were used to minimize T1 effects. To estimate T2* effects, three and seven serial echoes were obtained at different echo times. Sequence parameters are summarized in Table 1.


Single-voxel MRS data were acquired from 30×30×30 mm voxels obliquely positioned by a radiologist (D.B.P. with 10 years of experience with liver imaging) on the Couinaud segment V that corresponded to the location of the biopsy. Localizer sequences at maximum inspiration, maximum expiration, and during free breathing were used to ensure the voxel was inside the liver during the entire respiratory cycle. Care was taken to avoid liver edges, large hepatic vessels, and the biliary tree. Water suppression was not performed and automated shimming generated water line widths of 40–50 Hz.

Point -resolved spectroscopy sequences (PRESS) were used for MRS acquisition. For T2 correction, a multi-echo version was used and 8 spectra were collected at echo times (TE) of 40, 50, 60, 70, 80, 90, 100, and 110 ms. To minimize T1 effects, the repetition time (TR) was 2000 ms. The number of data points collected was 1024 across a 2000 Hz spectral window and measurements were the averages of 2 acquisitions (Figure 1). To estimate the fat fraction, single-echo, single-voxel MRS data were acquired using the same sequence with the following parameters: a TR of 4000 ms, a TE of 40 ms, a spectral window of 2000 Hz, the number of data points collected was 1024, and the number of acquisitions was 40 (Figure 1 - detail).

Figure 1. Representative MRS data obtained from a 62-year-old woman with type 2 diabetes and moderate steatosis.

T2 estimation was done by fitting the multiecho dataset for both water and fat components considering the spectral modeling. Fat fraction was calculated as described in details throughout the text using the T2-corrected single echo datasets (shown in detail).

Gradient echo image analysis

All MR imaging results were interpreted by one radiologist (D.B.P. with 10 years of experience with liver imaging), who was blinded to spectroscopic results. Analysis was performed using a custom analysis package in Matlab (Mathworks, Natick, MA). The radiologist manually drew a region of interest (ROI) approximately 900 mm2 in area and placed it at segment V of the liver, in the same region where MRS measurements and biopsies were performed (Figure 2). The fat fraction (including correction for T2* decay and noise bias) was obtained from average ROI values in magnitude images using a multipeak spectral model for fat detection [29], [30].

Figure 2. Representative out-of-phase MR images from a 62 year-old woman with type 2 diabetes and moderate steatosis.

MR images obtained from breath-hold T1-weighted triple-echo spoiled gradient-echo sequence. ROI was manually drawn at the spectroscopic voxel location (segment V, colocalized with liver biopsy), as shown. ROI, region of interest.

MRS analysis

MRS data were analyzed by a MR physicist (F.F.P., 5 years of experience with MRS) who was blinded to imaging results. Analysis used the Advanced Method for Accurate, Robust, and Efficient Spectral (AMARES) algorithm included in jMRUI [31]. At each echo time, the water (4.7 ppm) and fat (0.5–3 ppm) peaks were measured. Each measurable peak area was individually corrected for T2 decay using nonlinear least-square fits to determine their relative proton density. The relative proton densities of the fat peaks located underneath the water peaks were determined according to Hamilton et al. [32]. The total fat proton density was defined as the sum of all T2-corrected individual fat peaks. The proton density fat fraction (PDFF) was calculated as the ratio of the fat proton density to the sum of the fat and water proton densities.

Liver Biopsy and Evaluation

Liver specimens were obtained from all patients. Subcostal liver biopsies of the segment V of the liver were performed under US guidance using a 16-gauge Menghini biopsy needle. Specimens were 2 cm in length or longer, fixed in 10% formaldehyde solution, and then embedded in paraffin. The sections were then stained with hematoxylin and eosin, Masson's trichrome, and Perls' Prussian blue.

All biopsy slides were examined prospectively by a pathologist (V.L., 28 years of experience). Semiquantitative analysis of steatosis assessed the entire liver fragment and was performed by calculating the percentage of steatotic hepatocytes in the liver parenchyma. Results were classified into the following 4 groups according to the NASH Clinical Research Network (NASH-CRN) scoring system [33]: normal, <5%; mild, 5–33%; moderate, 33–66%; and severe,>66%. Steatosis measurements were scored in increments of 5% for more precise assessment. The pathologist's diagnosis of steatohepatitis was based on the presence of steatosis, hepatocyte injury (ballooning), and lobular inflammation. Fibrosis was scored from F0 to F4 [33]. The presence of siderosis was also evaluated.

Statistical Analyses

Statistical analyses were performed in R-project (version 2.15). Spearman statistic was used to estimate correlations. Linear regression analyses were conducted to explore the influence of steatohepatitis and fibrosis in liver fat measurements. For the linear models the areas under ROC curve were estimated with the Obuchowski's method[34]. Areas under the ROC curves were estimated by the trapezoidal method for mild (5–33%), moderate (33–66%), and severe steatosis (>66%) as reference values in biopsy evaluations. Decision thresholds were estimated for each MR technique through the maximization of the Youden J index on smoothed robust ROC curves. The ROC curves were compared using a variety of methods with a two-sided test. P values <0.05 indicated significant differences.



A total of 177 patients were interviewed for this study. Ninety-seven patients were excluded: 37 had decompensated cardiopulmonary or renal failure, 29 had hepatitis C, 20 had alcoholism, 7 had other chronic liver diseases, and 4 had coagulopathies. Thus, 80 patients underwent MRI. Five patients were subsequently excluded: 4 refused liver biopsy and 1 had precordial pain that precluded biopsy. Then, two patients were withdrawal: one due to an insufficient amount of biopsied tissue and 1 for granulomatous hepatitis upon histological analysis (Figure 3). Therefore, the final population of the study included 73 patients (60 women [82%] and 13 men [18%]) aged 54±9 years. The mean body mass index (BMI) was 31.4 kg/m2 (range, 23.2–42.7 kg/m2). Ninety-six percent (70 of 73) of patients were overweight (BMI>25 kg/m2) and 62% (45 of 73) were obese (BMI>30 kg/m2). MRI and MRS were performed during the same session and within 3 months after the liver biopsy. Thirty-six percent of all patients had elevated aminotransferase levels. Detailed demographic characteristics of the patients are included at Table 2.

Histological Analysis

Upon histology, the prevalence of steatosis (>5%) was 92% (67 patients). Mild steatosis was detected in 35 patients, moderate in 11 patients, and severe in 21 patients. Thirty-seven percent of the patients had steatohepatitis. Fibrosis was observed in 21 (29%) patients. The breakdown of NAFLD fibrosis scores among these patients were: F1, 12 patients; F2, 6 patients; F3, 2 patients; and F4, 1 patient. Two patients had mild siderosis.


Figure 4 shows correlations of the triple- and multi-echo sequences and MR spectroscopy with histopathology. Both MRI and MRS strongly correlated with liver biopsy findings (triple-echo: rho = 0.819, p<0.001; multi-echo: rho = 0.773, p<0.001; MRS: rho = 0.767, p<0.001). The correlation of PDFF in mild or moderate hepatic steatosis was found to be better than in severe steatosis.

Figure 4. Correlations between triple- and multi-echo sequences and MR spectroscopy versus histopathology examination.

Diagnostic Accuracy

The capability of each MR technique to diagnose steatosis was evaluated using histopathology for comparison. The sensitivities, specificities, thresholds, and AUC of each MR technique for mild, moderate, and severe steatosis are shown in figure 5. ROC curves were compared and a significant difference was observed for mild steatosis between multi-echo MRI and MRS (p = 0.033). However, no significant difference between triple-echo and multi-echo MRI (p = 0.093) or between triple-echo MRI and MRS (p = 0.42) was observed for mild steatosis. There was also no significant difference among the different MR methods for the ROC curves in moderate and severe steatosis.

Figure 5. Diagnostic performance for triple- and multi-echo sequences, and MRS using histopathology as the gold standard.

The best cut-off point was identified using the Youden index.

The median fat fraction determined by the different MR methods for mild steatosis varies from 4.98 to 7.06, for moderate steatosis from 10.84 to 14.27, and for severe steatosis from 18.26 to 27.75, and a progressive increase in the interquartile ranges using the same MR method is noted with no overlap (Table 3). Comparative analyses were performed in all groups, analyzed as paired comparisons between two groups to find out the differences between groups. All the comparative analyses showed p<0.05, suggesting that all the groups are different.

Table 3. Triple- and Multi-echo MRI and MRS medians for no steatosis, mild, moderate, and severe steatosis using biopsy evaluations as reference values.

One may see that the overall performance measures of the linear models do not change much when a univariable (with imaging only) and multivariable (with steatohepatitis and fibrosis (F0 vs. F1–F4)) are conducted in any of the MR technique (Table 4 and Table 5). Therefore, despite they are significant in the model, fibrosis and steatohepatitis contribute very little to the imaging results.

Table 4. Univariable linear regressions analyses for each one of the imaging techniques.

Table 5. Multivariable linear regressions analysis for each one of the imaging techniques.


Our study evaluated two MRI chemical shift techniques (triple- and multi-echo imaging) and MRS to diagnose steatosis and quantify fat in the livers of patients with type 2 diabetes. We compared these techniques to histopathological findings. Our results demonstrated that all MR techniques had high sensitivity and specificity for NAFLD diagnosis and strongly correlated with histopathology. In mild steatosis, a similar performance was observed between triple-echo sequences and MRS. In moderate and severe steatosis there was no significant difference among triple- and multi-echo sequences and MRS.

Although steatosis measurements by MR and histology were correlated, the actual values were different because the parameters used to estimate liver fat are different. Pathologists quantify the percentage of hepatocytes containing fat, whereas MR measures the signal from fat protons. Idilman et al. [35] found that PDFF better correlated with mild or moderate hepatic steatosis than severe steatosis, which is consistent with our data (Figure 4). These results suggest that hepatocytes in severe steatosis contain varied quantities of fat, but are not distinguished from each other upon histopathology (each is counted as a fat-containing hepatocyte). However, MR methods quantify the amount of fat in tissue. The fat fraction can be>2 times greater by histopathological assessment than by MR techniques, suggesting that many hepatocytes contain little fat. However, MRI and MRS measurements are similar [36], [37] to each other. Studies with larger numbers of patients should be performed to better understand the relationship between fat measurements by MR and histopathology.

Few studies have evaluated the influence of steatohepatitis or fibrosis in liver fat quantification [35], [38]. Although Idilman et al. [45]. found that hepatic fibrosis reduced the correlation between biopsy results and liver fat fraction, our results are similar to the ones of Tang et al. [38], where the presence of steatohepatitis and/or fibrosis did not affect MR steatosis measurements.

Our threshold values for the detection of steatosis were similar to those in other studies that used histopathology [36], [37], [39][42], but the appropriate thresholds have yet to be defined. In our study, the threshold for the detection of mild steatosis varied from 1.71 using MRS, to 4.09 using triple-echo sequence, to 7.53 using multi-echo MRI. Only a small overlap was observed between patients with no steatosis vs. mild steatosis for multi-echo sequence, where the IQR 75 for no steatosis was 6.24 and the IQR 25 for mild steatosis was 5.76. Neither other MR method nor other group had any overlap, suggesting that the technique may be able to accurately differentiate the patient groups. In addition, there were proportional increasing fat fraction medians for all MR techniques in patients with higher steatosis grades determined by histopathology. Thus, we believe that triple-echo, multi-echo, and MR spectroscopy are accurate non-invasive imaging methods for the diagnosis, grading, and follow-up of NAFLD.

Hepatic siderosis distorts the local magnetic field and causes loss of phase coherence and T2* shortening, which leads to signal loss at in-phase times compared to out-of-phase times [43], [44]. The T2* relaxation time varies widely among patients and using theoretical T2* relaxation times may lead to incorrect interpretations [45]. The T2* relaxation time can be calculated for each patient using triple- and multi-echo MRI, and with T2* correction, may prevent misinterpretation in cases of hepatic siderosis, a condition that can be associated with NAFLD [46]. In this study, only 2 patients had hepatic siderosis. Thus, we could not evaluate its effects on liver fat quantification.

Gradient-echo MRI appears to be the most appropriate for routine clinical use. In moderate and severe steatosis, no significant difference was observed for all MR methods. In our study, triple-echo sequence showed excellent correlation to histopathology and was similar to MRS to detect and quantify the liver fat fraction in mild steatosis. In addition, gradient-echo MR imaging can measure the liver fat fraction throughout the liver, has a short acquisition time, can be performed as a breath-hold sequence, and does not require hardware upgrades. Furthermore, it allows individual T2* corrections and the post-processing is easier, faster, and less prone to misinterpretation than MRS.

Although MRS enables highly accurate liver fat quantification, it requires strong magnetic fields, data processing is more complex and time consuming, it requires a large team highly specialized personnel, and is much more prone to errors. Lastly, it evaluates only a single voxel, which can lead to misinterpretation in cases of heterogeneous steatosis.

This study has some limitations. First, MRS was performed using a PRESS sequence, which may be more sensitive to J coupling effects [47]. This could lead to a systematic underestimation of fat relaxation times and slightly overestimate PDFF. Also, data acquisition time could be reduced by employing a single signal acquisition. We averaged many MRS acquisitions to ensure high signal-to-noise ratios and minimize baseline variations. Therefore, MRS had to be performed during free breathing where voxels moved 2–3 cm in the longitudinal plane. To minimize motion-related bias, the protocol included a coronal sequence during maximal inspiration and another during maximal expiration to ensure voxels remained properly positioned in the liver during acquisition. The MRI and MRS analyses were performed only in segment V, in the location corresponding to the liver biopsy and other liver segments were not evaluated to maintain clinically reasonable scan times. MRS voxels could also be difficult to position because of the biliary tree, vessels, and magnetic susceptibility in other liver segments. The prevalence of NAFLD in our population was extremely high and this could have negatively influenced the specificity, but this is inherent in the studied population. To the best of our knowledge, this is the first study to compare fat quantification by MRI and histopathology in diabetic patients. In addition, our study evaluated the performance of triple- and multi-echo MRI and MRS, with individual T2* and T2 corrections, respectively, which are important steps to allow comparisons between different magnets at different institutions.

In summary, MR measurements may be useful to screen for NAFLD. Liver fat quantification by MRI and MRS strongly correlated with histopathology, even in the presence of steatohepatitis and fibrosis. Triple- and multi-echo MRI and MRS were highly accurate and may be used as a substitute for liver biopsy. Due to the wide availability and easier post-processing, gradient-echo MRI appears to be the most appropriate for routine clinical use in the evaluation of NAFLD in diabetic patients.


Debora Lima; Adriana Passos; Erica Samão; D′Or Institute for Research and Education; FAPERJ.

Author Contributions

Conceived and designed the experiments: DBP RSR FFP JAON RMP. Performed the experiments: DBP FFP JAON VL CFFC RMP. Analyzed the data: DBP RSR FFP JAON LMS VL CFFC PEAAB RMP. Contributed reagents/materials/analysis tools: DBP RSR FFP JAON LMS VL CFFC AEA PEAAB PG RMP. Wrote the paper: DBP RSR FFP JAON LMS VL CFFC AEA PEAAB PG MBG RMP.


  1. 1. Farrell GC, Larter CZ (2006) Nonalcoholic fatty liver disease: from steatosis to cirrhosis. Hepatology 43: S99–S112.
  2. 2. Ong JP, Younossi ZM (2007) Epidemiology and natural history of NAFLD and NASH. Clin Liver Dis 11: 1–16, vii.
  3. 3. Szczepaniak LS, Nurenberg P, Leonard D, Browning JD, Reingold JS, et al. (2005) Magnetic resonance spectroscopy to measure hepatic triglyceride content: prevalence of hepatic steatosis in the general population. Am J Physiol Endocrinol Metab 288: E462–468.
  4. 4. Angulo P (2002) Nonalcoholic fatty liver disease. N Engl J Med 346: 1221–1231.
  5. 5. Browning JD, Szczepaniak LS, Dobbins R, Nuremberg P, Horton JD, et al. (2004) Prevalence of hepatic steatosis in an urban population in the United States: impact of ethnicity. Hepatology 40: 1387–1395.
  6. 6. Brunt EM, Tiniakos DG (2010) Histopathology of nonalcoholic fatty liver disease. World J Gastroenterol 16: 5286–5296.
  7. 7. Tiniakos DG, Vos MB, Brunt EM (2010) Nonalcoholic fatty liver disease: pathology and pathogenesis. Annu Rev Pathol 5: 145–171.
  8. 8. Charlton MR, Burns JM, Pedersen RA, Watt KD, Heimbach JK, et al. (2011) Frequency and outcomes of liver transplantation for nonalcoholic steatohepatitis in the United States. Gastroenterology 141: 1249–1253.
  9. 9. Leite NC, Salles GF, Araujo AL, Villela-Nogueira CA, Cardoso CR (2009) Prevalence and associated factors of non-alcoholic fatty liver disease in patients with type-2 diabetes mellitus. Liver Int 29: 113–119.
  10. 10. Cusi K (2009) Nonalcoholic fatty liver disease in type 2 diabetes mellitus. Curr Opin Endocrinol Diabetes Obes 16: 141–149.
  11. 11. Williams CD, Stengel J, Asike MI, Torres DM, Shaw J, et al. (2011) Prevalence of nonalcoholic fatty liver disease and nonalcoholic steatohepatitis among a largely middle-aged population utilizing ultrasound and liver biopsy: a prospective study. Gastroenterology 140: 124–131.
  12. 12. Targher G, Bertolini L, Padovani R, Rodella S, Tessari R, et al. (2007) Prevalence of nonalcoholic fatty liver disease and its association with cardiovascular disease among type 2 diabetic patients. Diabetes Care 30: 1212–1218.
  13. 13. Loomba R, Abraham M, Unalp A, Wilson L, Lavine J, et al. (2012) Association between diabetes, family history of diabetes and risk of nonalcoholic steatohepatitis and fibrosis. Hepatology
  14. 14. Neuschwander-Tetri BA, Clark JM, Bass NM, Van Natta ML, Unalp-Arida A, et al. (2010) Clinical, laboratory and histological associations in adults with nonalcoholic fatty liver disease. Hepatology 52: 913–924.
  15. 15. Silverman JF, O′Brien KF, Long S, Leggett N, Khazanie PG, et al. (1990) Liver pathology in morbidly obese patients with and without diabetes. Am J Gastroenterol 85: 1349–1355.
  16. 16. Wanless IR, Lentz JS (1990) Fatty liver hepatitis (steatohepatitis) and obesity: an autopsy study with analysis of risk factors. Hepatology 12: 1106–1110.
  17. 17. Harrison SA, Oliver D, Arnold HL, Gogia S, Neuschwander-Tetri BA (2008) Development and validation of a simple NAFLD clinical scoring system for identifying patients without advanced disease. Gut 57: 1441–1447.
  18. 18. Lazo M, Clark JM (2008) The epidemiology of nonalcoholic fatty liver disease: a global perspective. Semin Liver Dis 28: 339–350.
  19. 19. Gupte P, Amarapurkar D, Agal S, Baijal R, Kulshrestha P, et al. (2004) Non-alcoholic steatohepatitis in type 2 diabetes mellitus. J Gastroenterol Hepatol 19: 854–858.
  20. 20. Adams LA, Harmsen S, St Sauver JL, Charatcharoenwitthaya P, Enders FB, et al. (2010) Nonalcoholic fatty liver disease increases risk of death among patients with diabetes: a community-based cohort study. Am J Gastroenterol 105: 1567–1573.
  21. 21. Ma X, Holalkere NS, Kambadakone RA, Mino-Kenudson M, Hahn PF, et al. (2009) Imaging-based quantification of hepatic fat: methods and clinical applications. Radiographics 29: 1253–1277.
  22. 22. Qayyum A, Chen DM, Breiman RS, Westphalen AC, Yeh BM, et al. (2009) Evaluation of diffuse liver steatosis by ultrasound, computed tomography, and magnetic resonance imaging: which modality is best? Clin Imaging 33: 110–115.
  23. 23. Reeder SB, Cruite I, Hamilton G, Sirlin CB (2011) Quantitative Assessment of Liver Fat with Magnetic Resonance Imaging and Spectroscopy. J Magn Reson Imaging 34 spcone.
  24. 24. Strauss S, Gavish E, Gottlieb P, Katsnelson L (2007) Interobserver and intraobserver variability in the sonographic assessment of fatty liver. AJR Am J Roentgenol 189: W320–323.
  25. 25. Pickhardt PJ, Jee Y, O′Connor SD, del Rio AM (2012) Visceral adiposity and hepatic steatosis at abdominal CT: association with the metabolic syndrome. AJR Am J Roentgenol 198: 1100–1107.
  26. 26. Cassidy FH, Yokoo T, Aganovic L, Hanna RF, Bydder M, et al. (2009) Fatty liver disease: MR imaging techniques for the detection and quantification of liver steatosis. Radiographics 29: 231–260.
  27. 27. Schwenzer NF, Springer F, Schraml C, Stefan N, Machann J, et al. (2009) Non-invasive assessment and quantification of liver steatosis by ultrasound, computed tomography and magnetic resonance. J Hepatol 51: 433–445.
  28. 28. Reeder SB, Robson PM, Yu H, Shimakawa A, Hines CD, et al. (2009) Quantification of hepatic steatosis with MRI: the effects of accurate fat spectral modeling. J Magn Reson Imaging 29: 1332–1339.
  29. 29. Yokoo T, Bydder M, Hamilton G, Middleton MS, Gamst AC, et al. (2009) Nonalcoholic fatty liver disease: diagnostic and fat-grading accuracy of low-flip-angle multiecho gradient-recalled-echo MR imaging at 1.5 T. Radiology 251: 67–76.
  30. 30. Kuhn JP, Hernando D, Munoz del Rio A, Evert M, Kannengiesser S, et al. (2012) Effect of multipeak spectral modeling of fat for liver iron and fat quantification: correlation of biopsy with MR imaging results. Radiology 265: 133–142.
  31. 31. Stefan D, Di Cesare F, Andrasescu A, Popa E, Lazariev A, et al. (2009) Quantitation of magnetic resonance spectroscopy signals: the jMRUI software package. Measurement Science & Technology 20
  32. 32. Hamilton G, Yokoo T, Bydder M, Cruite I, Schroeder ME, et al. (2011) In vivo characterization of the liver fat (1)H MR spectrum. NMR Biomed 24: 784–790.
  33. 33. Kleiner DE, Brunt EM, Van Natta M, Behling C, Contos MJ, et al. (2005) Design and validation of a histological scoring system for nonalcoholic fatty liver disease. Hepatology 41: 1313–1321.
  34. 34. Obuchowski NA (2006) An ROC-type measure of diagnostic accuracy when the gold standard is continuous-scale. Stat Med 25: 481–493.
  35. 35. Idilman IS, Aniktar H, Idilman R, Kabacam G, Savas B, et al. (2013) Hepatic steatosis: quantification by proton density fat fraction with MR imaging versus liver biopsy. Radiology 267: 767–775.
  36. 36. d′Assignies G, Ruel M, Khiat A, Lepanto L, Chagnon M, et al. (2009) Noninvasive quantitation of human liver steatosis using magnetic resonance and bioassay methods. Eur Radiol 19: 2033–2040.
  37. 37. Hussain HK, Chenevert TL, Londy FJ, Gulani V, Swanson SD, et al. (2005) Hepatic fat fraction: MR imaging for quantitative measurement and display–early experience. Radiology 237: 1048–1055.
  38. 38. Tang A, Tan J, Sun M, Hamilton G, Bydder M, et al. (2013) Nonalcoholic fatty liver disease: MR imaging of liver proton density fat fraction to assess hepatic steatosis. Radiology 267: 422–431.
  39. 39. van Werven JR, Marsman HA, Nederveen AJ, Smits NJ, ten Kate FJ, et al. (2010) Assessment of hepatic steatosis in patients undergoing liver resection: comparison of US, CT, T1-weighted dual-echo MR imaging, and point-resolved 1H MR spectroscopy. Radiology 256: 159–168.
  40. 40. Kim H, Taksali SE, Dufour S, Befroy D, Goodman TR, et al. (2008) Comparative MR study of hepatic fat quantification using single-voxel proton spectroscopy, two-point dixon and three-point IDEAL. Magn Reson Med 59: 521–527.
  41. 41. Lee SS, Park SH, Kim HJ, Kim SY, Kim MY, et al. (2010) Non-invasive assessment of hepatic steatosis: prospective comparison of the accuracy of imaging examinations. J Hepatol 52: 579–585.
  42. 42. Noworolski SM, Lam MM, Merriman RB, Ferrell L, Qayyum A (2012) Liver Steatosis: Concordance of MR Imaging and MR Spectroscopic Data with Histologic Grade. Radiology 264: 88–96.
  43. 43. Westphalen AC, Qayyum A, Yeh BM, Merriman RB, Lee JA, et al. (2007) Liver fat: effect of hepatic iron deposition on evaluation with opposed-phase MR imaging. Radiology 242: 450–455.
  44. 44. O′Regan DP, Callaghan MF, Wylezinska-Arridge M, Fitzpatrick J, Naoumova RP, et al. (2008) Liver fat content and T2*: simultaneous measurement by using breath-hold multiecho MR imaging at 3.0 T–feasibility. Radiology 247: 550–557.
  45. 45. Thomsen C, Becker U, Winkler K, Christoffersen P, Jensen M, et al. (1994) Quantification of liver fat using magnetic resonance spectroscopy. Magn Reson Imaging 12: 487–495.
  46. 46. Mashhood A, Railkar R, Yokoo T, Levin Y, Clark L, et al. (2012) Reproducibility of hepatic fat fraction measurement by magnetic resonance imaging. J Magn Reson Imaging
  47. 47. Hamilton G, Middleton MS, Bydder M, Yokoo T, Schwimmer JB, et al. (2009) Effect of PRESS and STEAM sequences on magnetic resonance spectroscopic liver fat quantification. J Magn Reson Imaging 30: 145–152.