Differences in multi-echo chemical shift encoded MRI proton density fat fraction estimation based on multifrequency fat peaks selection in non-alcoholic fatty liver disease patients

Highlights

• Severe steatosis is common among subjects with metabolic syndrome.

• Iron overload is often present and related to severe steatosis in NAFLD patients.

• Simplified multifrequency MECSE methods underestimates steatosis quantification.

• PDFF-MECSE_p123456 is essential for an accurate estimation of liver steatosis.

AIM

To compare the performance of multi-echo chemical-shift-encoded (MECSE) magnetic resonance imaging (MRI) proton density fat fraction (PDFF) estimation, considering three different fat frequency peak combinations, for the quantification of steatosis in patients with non-alcoholic fatty liver disease (NAFLD).

MATERIALS AND METHODS

The present study was a prospective cross-sectional research of 121 patients with metabolic syndrome and evidence of hepatic steatosis on ultrasound, who underwent a 3 T MRI examination. All patients were studied with a multifrequency MECSE sequence. The PDFF was calculated using six peaks (MECSE_p123456), three peaks (MECSE_p456), and a single peak (MECSE_p5) model. The two simpler fat peak models were compared to the six peaks model, which was considered the reference standard. Linearity was evaluated using linear regression while agreement was described using Bland–Altman analysis.

RESULTS

The mean age was 47 (±9) years and BMI was 29.9 (±2.9) kg/m². Steatosis distribution was 15%/31%/54% (S1/S2/S3, respectively). Compared to MECSE_p123456, both models provided linear PDFF measurements (R₂= 0.99 and 0.97, MECSE_p456 and MECSE_p5 respectively). Regression slope (0.92; p<0.001) and mean Bland–Altman bias (–1.5%; 95% limits of agreement: –3.19%, 0.22%) indicated minimal underestimation by using PDFF-MECSE_p456. Nonetheless, mean differences in PDFF estimations varied from –1.5% (MECSE_p456, p=0.006) to –2.2% (MECSE_p5, p<0.001) when compared to full six fat frequencies model.

CONCLUSION

Although simpler spectral fat MECSE analysis shows a linear relationship with the standard six peaks model, their variation in estimated PDFF values introduces a low but clinically significant bias in fat quantification and steatosis grading in NAFLD patients.

Introduction

Hepatic steatosis is present in patients with a wide spectrum of chronic liver diseases, including alcoholic liver disease, non-alcoholic fatty liver disease (NAFLD), and viral hepatitis.^1,2 The burden of NAFLD in parallel to metabolic syndrome, with an estimated global prevalence of 24%, represents a public health problem and a serious concern.³ NAFLD, which comprises a spectrum from simple steatosis to steatohepatitis with varying degrees of fibrosis and iron overload, is the most common chronic liver disease (CLD) in developed countries.^4,5 Well-known risk factors for this condition are obesity, diabetes mellitus, hypertension, dyslipidaemia, and cardiovascular events.⁶ In addition, non-alcoholic steatohepatitis (NASH) is related to poor clinical outcomes as it can progress to cirrhosis and hepatocellular carcinoma.^7,8

Liver biopsy is considered the traditional reference standard technique to evaluate liver disease, staging from steatosis to steatohepatitis and grading iron deposition and fibrosis. Unfortunately, this is an invasive and painful technique, refused by many patients, associated with complications, and prone to sampling error due to the heterogeneity of liver diseases.^9,10 Furthermore, liver biopsy cannot be ethically considered as a screening or a follow-up diagnostic technique in these patients.

Non-invasive imaging-based alternatives to liver biopsy are accepted in daily clinical practice.¹¹ Magnetic resonance imaging (MRI) plays an essential role in the diagnosis and grading of hepatic steatosis, being able to provide accurate information of the whole organ and recently considered a precise “virtual biopsy” for liver fat quantification.^1,2,12 In addition, MRI quantitative parameters allow detection of relevant liver changes before the patient is symptomatic, as an opportunistic screening method anticipating advanced disease stages. Therefore, the European Association for the Study of the Liver (EASL) Clinical Practice Guidelines defines that NAFLD can be diagnosed with quantitative MRI assessment, without histological confirmation.⁶

Several MRI chemical shift techniques are used for the quantification of fat content in NAFLD, including dual-echo signal fat fraction, multi-echo proton density fat fraction (PDFF), and spectroscopy (MRS). Although single-voxel MRS has traditionally been proposed as the reference technique for liver fat quantification, it suffers from a clear subsample bias not considering the heterogeneous distribution of fat and iron within the liver parenchyma.² This limitation can be overcome by the use of multi-echo chemical shift-based encoded gradient echo (MECSE) magnitude and phase complex approaches, providing reliable and fast PDFF with spatially resolved quantification of liver triglycerides, when confounding factors are avoided.^13,14 In addition, MRI-PDFF measurements, with a corrected MECSE full multifrequency approach, has a close linear correlation to MRS with a mean difference smaller than 1% and small variability.^12,15,16 Consequently, nowadays the MECSE techniques are recommended for the non-invasive quantification of steatosis in patients with NAFLD.^1,6,12,17,18 The MECSE sequence allows simultaneous estimation of the fraction of mobile protons related to fat, as PDFF measurements corrected for R2∗, mainly influenced by the presence of tissue iron, minimising T1 influence, mainly related to fat and water different relaxation times.19, 20, 21, 22, 23 The spectral complexity of fat signal relates to the distinct proton species within the triglyceride components, and at least six constituents can be observed in clinical standard of care MRI acquisitions due to the chemical frequency shift (Fig 1). The six principal distinct spectral peaks at different resonance frequencies are 5.48, 4.20, 2.75, 2.26, 1.43, and 0.90 ppm; although some authors used up to 10 triglyceride components.²⁴ To calculate the total PDFF, the relative amplitudes of these six main peaks have to be considered, their approximate proportions are 4.7, 3.9, 0.6, 12, 70 and 8.8%, respectively. Although accurate PDFF quantification should be obtained using all six main fat peaks (MECSE_p123456),^2,25, 26, 27 peaks 1 to 3 (5.48, 4.20, 2.75 ppm) have smaller overall contribution and are closest to the water peak, which might introduce quantifying errors.²⁰ Therefore, some authors have used different multi-interference combinations with fewer fat peaks to simplify the MECSE method. Other published methods used to calculate the MRI-derived PDFF from the same complex source data included the three peaks close to the main methylene fat resonance (2.26, 1.43, 0.90 ppm; MECSE_p456)19, 20, 21^,28, 29, 30, 31 and the main methylene fat peak (1.43 ppm; MECSE_p5).^32,33

The aim of the present study was to compare the differences in non-invasive diagnosis of liver steatosis when using different MECSE fat peaks combinations in a large series of patients with metabolic syndrome, using the complex-based multifrequency six peaks as the reference standard. It was hypothesised that PDFF estimations performed with the simpler MECSE models might introduce a quantification bias that entails different clinical results when diagnosing and staging steatosis.