A comparison of liver fat content as determined by magnetic resonance imaging-proton density fat fraction and MRS versus liver histology in non-alcoholic fatty liver disease

Abstract

Background

Many imaging methods have been defined for quantification of hepatic steatosis in non-alcoholic fatty liver disease (NAFLD). However, studies comparing the efficiency of magnetic resonance imaging-proton density fat fraction (MRI-PDFF), magnetic resonance spectroscopy (MRS), and liver histology for quantification of liver fat content are limited.

Purpose

To compare the efficiency of MRI-PDFF and MRS in the quantification of liver fat content in individuals with NAFLD.

Material and Methods

A total of 19 NAFLD patients underwent MRI-PDFF, MRS, and liver biopsy for quantification of liver fat content. The MR examinations were performed on a 1.5 HDx MRI system. The MRI protocol included T1-independent volumetric multi-echo gradient-echo imaging with T2* correction and spectral fat modeling and MRS with STEAM technique.

Results

A close correlation was observed between liver MRI-PDFF- and histology- determined steatosis (r = 0.743, P < 0.001) and between liver MRS- and histology-determined steatosis (r = 0.712, P < 0.001), with no superiority between them (ƶ = 0.19, P = 0.849). For quantification of hepatic steatosis, a high correlation was observed between the two MRI methods (r = 0.986, P < 0.001). MRI-PDFF and MRS accurately differentiated moderate/severe steatosis from mild/no hepatic steatosis (P = 0.007 and 0.013, respectively), with no superiority between them (AUC_MRI-PDFF = 0.881 ± 0.0856 versus AUC_MRS = 0.857 ± 0.0924, P = 0.461).

Conclusion

Both MRI-PDFF and MRS can be used for accurate quantification of hepatic steatosis.

Keywords

Non-alcoholic fatty liver disease magnetic resonance imaging (MRI)proton density fat fraction (PDFF)magnetic resonance spectroscopy (MRS)liver biopsy

Introduction

Non-alcoholic fatty liver disease (NAFLD) affects approximately 10–30% of the general population with an ever increasing prevalence in accordance with the obesity epidemic (1 –3). Most patients with NAFLD have no symptoms or signs and are identified incidentally on routine blood tests or abdominal imaging (3). However, in addition to this silent form, the disease can present with non-alcoholic steatohepatitis (NASH), cirrhosis, or even hepatocellular carcinoma (1 –3). While the current gold standard for the diagnosis of NAFLD is liver biopsy, it has several limitations including its cost, invasiveness, complications, sampling variability, and inter-observer discordance (4,5). Its invasiveness also presents a challenge for any follow-up evaluation. Imaging modalities such as ultrasonography, computed tomography (CT), and conventional magnetic resonance imaging (MRI) are frequently used in the diagnosis of NAFLD (6); however, none provides satisfactory quantitative data necessary for grading the disease and follow-up evaluation.

Magnetic resonance spectroscopy (MRS) can accurately quantify lipid fraction relative to water in the tissue and has been accepted as a reference imaging method for the assessment of liver fat content (7,8). The acquisition parameters, method of analysis and location of volume assessed can affect the accuracy of its evaluation. Proton density fat fraction (PDFF) calculation is a recent MRI technique which eliminates biases like T1 bias, T2^* decay, spectral complexity of fat, noise bias, and eddy currents in conventional MRI (9 –11). This technique enables estimation of PDFF of the liver, which has shown good correlation with MRS- (10 –14) and biopsy- (15,16) determined fat content in the liver.

Previous studies comparing the efficiency of liver MRI-PDFF, MRS, and liver histology in terms of quantification of liver fat content are limited (17). The aim of the present study was to determine the efficiency of two MRI methods – MRI-PDFF and MRS – in the quantification of liver fat content in individuals with NAFLD as compared with that determined by liver biopsy.

Material and Methods

Study population

A total of 19 consecutive individuals diagnosed with NAFLD in the Liver Diseases Outpatient Clinic, who underwent both MRI-PDFF and MRS for quantification of hepatic steatosis between June 2010 and December 2010, were included in this retrospective study. Diagnosis of NAFLD in each case was based on biochemical, radiological and histological criteria and on exclusion of other forms of acute and chronic liver diseases (1,2,18,19). Criteria for inclusion were: (i) age >16 years; (ii) evidence of absent or minimal alcohol consumption: <20 g alcohol/day for women and <30 g alcohol/day for men; (iii) absence of confounding disease including acute and/or chronic viral hepatitis (hepatitis A, B, or C); and (iv) exclusion of other forms of liver disease including autoimmune, drug-induced, and metabolic liver diseases. This study was approved by the institutional review board. All patients with NAFLD provided a written informed consent form for liver biopsy and MR examination.

Liver injury and function tests: Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma glutamyl transpeptidase (GGT), alkaline phosphatase (ALP), bilirubin, fasting glucose, cholesterol, triglycerides levels, and complete blood cell counts were measured by our central laboratory using standard reagents. Insulin was measured by radioimmunoassay. For exclusion of other forms of liver disease, serological markers for viral infections, serum iron, ferritin, copper, ceruloplasmin, and alpha-1 antitrypsin levels were measured, and serological studies for anti-nuclear antibody, anti-smooth muscle antibody and anti-mitochondrial antibodies were performed as mentioned in the previous study (16).

Histological assessments: All liver biopsy specimens were retrieved from the archives of the Department of Pathology. Biopsy specimens were evaluated by an experienced pathologist blinded to the clinical and biochemical data (16). First, the percentages of steatotic hepatocytes were documented. Biopsies were scored using the NASH Clinical Research Network (CRN) NAFLD Activity Score (NAS) and fibrosis score (20). Hepatic steatosis was classified from grade 0 to grade 3. Minimal steatosis was defined as <5% steatosis; mild steatosis, grade 1, 5–32%; moderate steatosis, grade 2, 33–65%; and severe steatosis, grade 3, as >66%. Lobular inflammation was evaluated at 200×optical field as: 0, no inflammation; 1, <2 foci; 2, 2–4 foci; and 3, >4 foci of inflammation. In addition to NASH CRN NAS, portal inflammation was evaluated as: 0, none; 1, mild; 2, moderate; and 3, marked. Fibrosis was scored in the range of 0–4, and significant fibrosis was defined as stage 2–4 fibrosis. NAS (score 0–8) was calculated based on the grade of steatosis (grade 0–3), lobular inflammation (grade 0–3), and ballooning (grade 0–2) (20).

Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Obesity was defined based on World Health Organization criteria with BMI of 25–29.9 kg/m² defining overweight and BMI ≥30 kg/m² defining obesity. Insulin resistance (IR) was calculated on the basis of fasting plasma glucose and insulin values using the homeostasis model assessment-insulin resistance method (HOMA-IR: plasma glucose (mg/dl)×insulin (µu/mL) / 405).

MR examinations

The MR examinations were performed on a 1.5 T HDxt MRI system (GE Healthcare, Milwaukee, WI, USA). An 8-channel phased array body coil was used for MR acquisitions. The subjects were examined in the supine position. A three-plane localization imaging gradient echo sequence was performed at the beginning of the examination. The MRI protocol included IDEAL-IQ (Iterative Decomposition of water and fat with Echo Asymmetry and Least square estimation) sequence and single voxel spectroscopy.

The IDEAL-IQ sequence was used to measure R2* (1/T2*) and PDFF in the liver simultaneously in a single acquisition. The details of the sequence have been described elsewhere in greater detail (9). The parameters of this sequence were: TR, 12.9 ms; field of view, 35–40 cm; matrix, 224 × 160; bandwidth, 125 kHz; slice thickness, 10 mm; and a single 3D slab with 22–28 slices was acquired. We acquired six different echoes in the range of 1.6–9.8 ms. The images were processed using the software provided by the manufacturer to create water, fat, IP, OP, R2*, and fat fraction maps. All measurements were corrected for heterogeneity using the homogeneity maps from auto calibration. This sequence was obtained in all patients while breath-holding for <25 s.

Image Processing: By using a workstation (AW 4.4, GE Healthcare), a radiologist who was unaware of the biopsy results placed an elliptic region of interest (ROI) of 4 cm² on the fat fraction images calculated from the IDEAL-IQ sequence at Couinaud segments V–VI.

Single-voxel spectra were measured using the Stimulated Echo Acquisition Mode (STEAM) pulse sequence (21,22). A 20 × 20 × 20 mm (8 cm³) MRS voxel was placed on the Couinaud segment V–VI of the liver, avoiding vessels and the edge of the liver. The STEAM pulse sequence parameters without water suppression were: TR, 3 s; TE, 12 ms; data points, 2048; acquisition time, 20 s; and number of averages, 6.

The raw spectra data were processed using a commercially available postprocessing software (Spectro Analysis by GE [SAGE]). The phase correction based on water was applied automatically. SAGE software was used to measure the peak value of water peak at 4.7 ppm and methylene peak (CH₂) at 1.2 ppm. The percentile hepatic lipid content or fat fraction were calculated as the methylene peak (CH2 Peak) divided by the sum of water peak and methylene peak and multiplied by 100 from segments V–VI.

Assessments

The correlations between histology-determined steatosis and both liver MRI-PDFF and liver MRS were assessed, and these correlations were compared in terms of superiority between methods. Estimation of liver fat content using MRI-PDFF and MRS in patients with/without fibrosis was analyzed. The accuracy of MRI-PDFF and MRS in terms of differentiation of moderate/severe hepatic steatosis from mild/no hepatic steatosis was assessed, and these accuracies were compared in terms of superiority between methods.

Statistical analysis

The degree of association between continuous variables was calculated by Pearson’s correlation coefficient. The difference between two correlation coefficients obtained from independent samples was assessed using z statistics (http://faculty.vassar.edu/lowry/rdiff.html). Comparison between two groups in terms of continuous variables was assessed by Student’s t test. Receiver operating characteristic (ROC) curves were used to describe and compare the performance of diagnostics value of the two radiological imaging methods. Cutoff ranges were calculated around the optimal cutoff to maximize sensitivity and specificity to differentiate moderate or severe steatosis (>33% steatotic hepatocytes) from mild or no hepatic steatosis (<33% steatotic hepatocytes). The area under the corresponding curves (AUC) was calculated and compared. For all tests, a two-tailed P value of <0.05 was considered statistically significant.

Results

At the time of the imaging, the mean age of the patients was 41.7 ± 14.6 years (M/F: 15/4). Mean±standard deviation of serum AST, ALT, GGT, and total bilirubin levels were 41.3 ± 15.5 U/L, 78.2 ± 42.4 U/L, 75.7 ± 63.4 U/L, and 0.8 ± 0.4 mg/dL, respectively. The characteristics of all patients with NAFLD are shown in Table 1. On the liver biopsy examination, 26.3% of the biopsy samples revealed grade 1 hepatic steatosis, 26.3% grade 2 hepatic steatosis, and 36.8% grade 3 hepatic steatosis. Necroinflammation, either lobular or portal, was graded as follows: 15.8% of the patients (3/19) had no necroinflammation and 84.2% (16/19) had grade 1 necroinflammation. Of note, 36.8% of the samples (7/19) revealed the presence of fibrosis. The histological features are shown in Table 2.

Table 1.

The characteristics of 19 patients with NAFLD.

Age (years)	41.7 ± 13.1
Gender (M/F)	15/4
BMI (kg/m²)	27.5 ± 3.3
Obesity (%)	37
Diabetes mellitus (%)	21
Insulin resistance (>2.7) (%)	68
Hypertension (%)	11
Hyperlipidemia (%)	53
Fasting glucose (N: 75–115 mg/dL)	88.1 ± 23.7
Cholesterol (N: 120–200 mg/dL)	210.7 ± 35.5
Triglyceride (N: 40–165 mg/dL)	151.8 ± 47.4
HDL (N: 40–60 mg/dL)	50 ± 11.8
Serum AST (N: 10–37 U/L)	41.3 ± 15.5
Serum ALT (N: 10–37 U/L)	78.2 ± 42.4
Serum GGT (N: 0–55 U/L)	75.7 ± 63.4
Serum total bilirubin (N: 0.2–1.3 mg/dL)	0.8 ± 0.41

Data are mean±standard deviation.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma glutamyl transpeptidase; HDL, high density lipoprotein.

Table 2.

The histopathological characteristics of 19 patients with NAFLD.

	n	%
Steatosis
Grade 0 (<5%)	2	10.5
Grade 1 (5–32%)	5	26.3
Grade 2 (33–65%)	5	26.3
Grade 3 (>66%)	7	36.8
Inflammation
A0	3	15.8
A1	16	84.2
A2	0	0
A3	0	0
Fibrosis
F0	12	63.2
F1	6	31.6
F2	1	5.2
F3	0	0

Lobular inflammation was evaluated at 200×optical field as: 0, no inflammation; 1, <2 foci; 2, 2–4 foci; and 3, >4 foci of inflammation.

Fibrosis was defined as: 0, none; 1, perisinusoidal or periportal; 2, perisinusoidal and portal/periportal; and 3, bridging fibrosis.

The median interval between the liver biopsy and MR examinations was 34 days (range, 0–257 days). Mean liver MRI-PDFF, MRS, and percentage of histological steatosis were 17.2% ± 9.4, 14.8% ± 9.4, and 48.1% ± 29.9, respectively. A high correlation was observed between liver MRI-PDFF- and histology-determined steatosis (r = 0.743, P < 0.001) and between liver MRS- and histology-determined steatosis (r = 0.712, P < 0.001) for quantification of hepatic steatosis (Figs 1 and 2a, b). However, no superiority between the two imaging methods was observed (ƶ = 0.19, P = 0.849). Fat fraction assessed by MRI-PDFF showed better correlation with histopathology in patients without fibrosis (n = 12, r = 0.88, P < 0.001) than in patients with fibrosis (n = 7, r = 0.45, P = 0.315), but without statistical significant difference (ƶ = 1.48, P = 0.139). Similar results were obtained for MRS (r of 0.88 and P < 0.001 for the former, r of 0.41 and P = 0.364 for the latter); however, without statistical significant difference (ƶ = 1.56, P = 0.119). For quantification of hepatic fat content, a high correlation was observed between the two MRI methods (r = 0.986, P < 0.001) (Fig. 3).

Fig. 1.

In-phase, opposed-phase, MRI-PDFF images, and MR spectra of two patients with NAFLD. Patient A’s (top row) liver MRI-PDFF, MRS, and percentage of histological steatosis were 16.3%, 14.2%, and 30%, respectively, whereas patient B’s (bottom row) liver MRI-PDFF, MRS, and percentage of histological steatosis were 21.6%, 18.5%, and 30%, respectively.

Fig. 2.

Scatterplot shows the correlation between liver MRI-PDFF- and histology-determined steatosis (r = 0.743, P < 0.001) (a) and between liver MRS- and histology-determined steatosis (r = 0.712, P < 0.001) (b).

Fig. 3.

Scatterplot shows the close correlation between liver MRI-PDFF and MRS (r = 0.986, P < 0.001).

In subgroup analyses, seven patients had none or mild hepatic steatosis whereas the remaining 12 patients had moderate and severe hepatic steatosis. Mean liver MRI-PDFF, MRS, and percentage of histological steatosis in patients with none or mild hepatic steatosis were 10% ± 6.5, 8% ± 6, and 13.4% ± 11.9, respectively. Mean liver MRI-PDFF, MRS, and percentage of histological steatosis in patients with moderate or severe hepatic steatosis were 21.4% ± 8.4, 18.7% ± 9, and 68.3 ± 13.4, respectively. MRI-PDFF and MRS accurately differentiated moderate/severe hepatic steatosis from mild/no hepatic steatosis (P = 0.007 and 0.013, respectively; Fig. 4a and b). The cutoff value for MRI-PDFF that discriminated moderate/severe steatosis from none/mild steatosis was 10% (AUC ± SE: 0.881 ± 0.086, 95% CI: 0.71–1.00, P = 0.007), with a sensitivity of 100%, specificity of 71.4%, positive predictive value of 85.7%, and negative predictive value of 100%. The cutoff value for MRS that discriminated moderate/severe steatosis from none/mild steatosis was 9% (AUC±SE: 0.857 ± 0.092, 95% CI: 0.68–1.00, P = 0.011), with a sensitivity of 92%, specificity of 71%, positive predictive value of 84.6%, and negative predictive value of 83.3%. A comparison of accuracy between these two methods revealed no superiority (AUC_MRI-PDFF = 0.881 ± 0.086 versus AUC_MRS = 0.857 ± 0.092, P = 0.461).

Fig. 4.

Boxplot shows the accuracy of liver MRI-PDFF (P = 0.005) (a) and MRS (P = 0.010) (b) in the differentiation of moderate and severe steatosis from mild or no hepatic steatosis.

Discussion

Ultrasonography remains the first method for evaluation of the presence and severity of NAFLD, with a low sensitivity and specificity. CT is the other imaging method for determining liver fat, which is based on X-ray penetration of the tissue. Simple measurement of liver attenuation in unenhanced images is considered for estimation of fatty liver. However, it cannot determine a certain value for steatosis beyond grading. Furthermore, patient exposure to radiation makes it unsuitable for use in the follow-up evaluation. MRS has been shown to accurately measure hepatic lipid fraction. However, it has several limited clinical applications. There have been several advances in chemical shift based water-fat separation methods in quantifying liver fat content in recent years. MRI-PDFF is a novel MRI technique, which addresses confounding factors for accurate lipid fraction quantification. It permits quantification of lipid fraction of the whole liver, which is a limitation for both MRS and liver biopsy.

There are many studies in the literature that have evaluated different MRI techniques to assess hepatic fat. Despite the positive strong correlation observed between different in-phase and out-of-phase gradient-echo MRI (23), older techniques have limited ability to quantify hepatic fat as they are prone to biases like T1 bias, T2^* decay, spectral complexity of fat, noise bias, and eddy currents (9 –11). In some of the previous studies, the techniques used were similar to those used in the present study, and a significant correlation between MRI-PDFF- and MRS-determined hepatic fat fraction was observed (10 –14). In our study, we observed a high correlation between liver MRI-PDFF and liver MRS (r = 0.986) in accordance with these previous studies. However, in these previous studies, the reference technique for quantification of liver fat was MRS, and the correlation of these techniques with liver biopsy- determined steatosis was not well evaluated. In the present study, we evaluated the efficiency of MRI-PDFF- and MRS-determined liver fat content in patients with NAFLD in comparison with liver biopsy-determined steatosis. For quantification of hepatic steatosis, a close correlation was observed among liver MRI-PDFF- and histology-determined steatosis (r = 0.743, P < 0.001) and between liver MRS- and histology-determined steatosis (r = 0.712, P < 0.001). No superiority between the two imaging methods was observed (P = 0.426). There is just one study available in the literature that compared liver MRI-PDFF-, MRS-, and histology-determined steatosis in patients with NAFLD before and after treatment (17). In this study, investigators observed a significant correlation between MRI-PDFF-, MRS-, and histology-determined steatosis at both baseline and follow-up. The findings of the present study are consistent with this previous study.

McPherson et al. demonstrated a significant correlation between MRI (Dixon in phase/out phase and with/without fat saturation), MRS and liver histology in terms of quantification of liver fat content (24). The investigators suggested that the accuracy of the MRI techniques was lower in patients with moderate and severe fibrosis, with a better accuracy in MRS. The present study confirms this finding and demonstrated that the estimation of liver fat content by both investigated MR imaging methods was better in patients with no hepatic fibrosis. As mentioned by McPherson et al., we think that the correlation between MRI methods- and histology-determined steatosis is lower in patients with fibrosis because of the reduction in the number of hepatocytes due to displacement by fibrosis (24). Unfortunately, the difference in correlations was not evaluated in the present study because of the small sample size of patients with fibrosis.

There are some limitations of this study. Because of the small sample size of hepatic inflammation and fibrosis, the potential confounding effect of these parameters on the relationship between histology-determined steatosis and per-patient MRI-PDFF and MRS could not be analyzed. The field strength of MR scanner was 1.5 T, which means a reduction in the chemical shift effect. However, it has been shown in a previous study that MRI-PDFF- determined liver fat fractions do not differ at different field strength scanners (25). In the present study, the MRS method used had a very short TE to minimize the T2 effect. However, T2 correction was not performed, which is important to improve the accuracy of fat fraction measurements especially when iron is present. We found a strong correlation between the two MRI techniques despite one of them (MRS) measuring a single fat peak. We believe that the underestimation of fat fraction with single peak MRS measurements is affected (or compensated) to some degree by the T2 relaxation effects, which can result in overestimation of fat fraction in the MRS method. However, considering the fact that the MRI method does include multiple corrections, we believe that the MRS method used here tends to underestimate the fat fraction, which is a limitation of the present study.

In conclusion, based on the results of the present study, both MRI-PDFF and MRS can be used for accurate quantification of hepatic steatosis and differentiation of moderate and severe steatosis from none or mild steatosis. No superiority between these two imaging methods was observed, but MRI-PDFF has the potential for total liver fat quantification.

Footnotes

Acknowledgements

We thank Serap Ersert, Ozlem Ozel, Elif Polat, and Nurhan Tazegun for valuable contributions to this study.

Conflict of interest

Azim Celik is employee of GE healthcare. His financial activities are not related to the present article.

Funding

Ramazan Idilman is an associate member of the TUBA. Musturay Karcaaltincaba has been supported by the Turkish Academy of Sciences (TUBA), in the framework of the Young Scientist Award Program (EA-TUBA-GEBIP/2011).

References

Angulo

. Nonalcoholic fatty liver disease. N Engl J Med 2002; 346: 1221–1231.

Neuschwander-Tetri

Caldwell

. Nonalcoholic steatohepatitis: summary of an AASLD single topic conference. Hepatology 2003; 37: 1202–1219.

Vuppalanchi

Chalasani

. Nonalcoholic fatty liver disease and nonalcoholic steatohepatitis: selected practical issues in their evaluation and management. Hepatology 2009; 49: 306–317.

Ratziu

Charlotte

Heurtier

. Sampling variability of liver biopsy in nonalcoholic fatty liver disease. Gastroenterology 2005; 128: 1898–1906.

Vuppalanchi

Unalp

Natta

. Increased diagnostic yield from liver biopsy in suspected NAFLD using multiple cores and multiple readings. Gastroenterology 2007; 132: 809–809.

Schwenzer

Springer

Schraml

. Non-invasive assessment and quantification of liver steatosis by ultrasound, computed tomography and magnetic resonance. J Hepatol 2009; 51: 433–445.

Szczepaniak

Nurenberg

Leonard

. Magnetic resonance spectroscopy to measure hepatic triglyceride content: prevalence of hepatic steatosis in the general population. Am J Physiol Endocrinol Metab 2005; 288: 462–468.

Cowin

Jonsson

Bauer

. Magnetic resonance imaging and spectroscopy for monitoring liver steatosis. J Mag Reson Imaging 2008; 28: 937–945.

Reeder

Robson

. Quantification of hepatic steatosis with MRI: the effects of accurate fat spectral modeling. J Mag Reson Imaging 2009; 29: 1332–1339.

10.

Yokoo

Bydder

Hamilton

. Nonalcoholic fatty liver disease: diagnostic fat-grading accuracy of low-flip-angle multiecho gradient-recalled-echo MR imaging at 1.5 T. Radiology 2009; 251: 67–76.

11.

Kim

Taksali

Dufour

. A comparative MR study of hepatic fat quantification using single-voxel proton spectroscopy, two-point Dixon and three-point IDEAL. Magn Reson Med 2008; 59: 521–527.

12.

Meisamy

Hines

Hamilton

. Quantification of hepatic steatosis with T1-independent, T2-corrected MR imaging with spectral modeling of fat: blinded comparison with MR spectroscopy. Radiology 2011; 256: 767–775.

13.

Hines

Frydrychowicz

Hamilton

. T(1) independent, T(2) (*) corrected chemical shift based fat-water separation with multi-peak fat spectral modeling is an accurate and precise measure of hepatic steatosis. J Magn Reson Imaging 2011; 33: 873–881.

14.

Kim

Nayak

. Comparison of fat-water MRI and single-voxel MRS in the assessment of hepatic and pancreatic fat fractions in humans. Obesity 2010; 18: 841–847.

15.

Tang

Tan

Sun

. Nonalcoholic fatty liver disease: MR imaging of liver proton density fat fraction to assess hepatic steatosis. Radiology 2013; 267: 422–431.

16.

Idilman

Aniktar

Idilman

. Hepatic steatosis: quantification by proton density fat fraction with MR imaging versus liver biopsy. Radiology 2013; 267: 767–775.

17.

Noureddin

Lam

Peterson

. Utility of magnetic resonance imaging versus histology for quantifying changes in liver fat in nonalcoholic fatty liver disease trials. Hepatology 2013; 58: 1930–1940.

18.

Saadeh

Younossi

Remer

. The utility of radiological imaging in nonalcoholic fatty liver disease. Gastroenterology 2002; 123: 745–750.

19.

Brunt

Janney

Di Bisceglie

. Nonalcoholic steatohepatitis: a proposal for grading and staging the histological lesions. Am J Gastroenterol 1999; 94: 2467–2474.

20.

Kleiner

Brunt

Van Natta

. Design and validation of a histological scoring system for nonalcoholic fatty liver disease. Hepatology 2005; 41: 1313–1321.

21.

Frahm

Bruhn

Gyngell

. Localized high-resolution proton NMR spectroscopy using stimulated echoes: initial applications to human brain in vivo. Magn Reson Med 1989; 9: 79–93.

22.

Reeder

Cruite

Hamilton

. Quantitative assessment of liver fat with magnetic resonance imaging and spectroscopy. J Mag Reson Imaging 2011; 34: 729–749.

23.

Ramalho

Herédia

de Campos

. In-phase and out-of-phase gradient-echo imaging in abdominal studies: intra-individual comparison of three different techniques. Acta Radiol 2012; 53: 441–449.

24.

McPherson

Jonsson

Cowin

. Magnetic resonance imaging and spectroscopy accurately estimate the severity of steatosis provided the stage of fibrosis is considered. J Hepatol 2009; 51: 389–397.

25.

Kang

Cruite

Shiehmorteza

. Reproducibility of MRI-determined PDFF across two different MR scanner platforms. J Magn Reson Imaging 2011; 34: 928–934.