Noninvasive Quantification of Liver Fat Content by Different Gradient Echo Magnetic Resonance Imaging Sequences in Patients with Nonalcoholic Fatty Liver Disease.

BACKGROUND: Noninvasive quantification of liver fat by gradient echo (GRE) technique is an interesting issue in quantitative magnetic resonance imaging. In this study, the fat content in patients with nonalcoholic fatty liver disease (NAFLD) was quantified with GRE sequences with different T 1 and T 2* weighting.
METHODS: This prospective, cross-sectional study was performed on thirty NAFLD patients. Sixteen GRE sequences with different T 1 weighting were performed with four echo times. In each sequence, repetition time (TR) or flip angle was changed and other parameters were fixed. Forty-eight fat indexes (FIs) from 16 sequences were calculated based on three methods. To determine the relationship between FIs and histological findings, Pearson's correlation coefficient was used at the level of 1% significance.
RESULTS: Mean FIs which obtained from Eq. 3 have the maximum values in comparison to other FIs. The maximum FI was 23.58%, which related to heavily T 1 weighted sequence obtained with method 3. The minimum FI was -2.49%, which related to the minimal T 1 weighted obtained with method 2. FIs increase with a flip angle, especially at low flip angles. Increase the TR parameter decrease the FIs gradually. Calculated FIs with methods 1 and 3 stronger correlated with histological findings relative to calculated FIs with method 2.
CONCLUSION: For fat quantification, T 1 relaxation effects probably more critical than T 2*. Flip angle parameter could be a major factor causing the overestimation of liver fat content. Sequences with low flip angle are more suitable for fat quantification with methods 1 and 3. In fat quantification with GRE techniques, it is possible that the third and fourth echoes are unnecessary.

Introduction

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disorder in Western countries. The prevalence rate of this disease is 20%–30% in the general population and 60%–80% among diabetics and obese patients.[12] Hepatic steatosis or accumulation of lipid vacuoles within hepatocytes is the primary histologic hallmark for diagnosis of NAFLD.[3]

Noninvasive detection and quantification of the steatosis has a considerable importance in clinical hepatology. In liver surgery, since the presence of moderate or severe steatosis may lead to graft failure in recipients and postponed the recovery process of donors, so the degree of steatosis should be measured accurately for transplant decision-making.[45] The steatosis should be carefully determined in NAFLD patients who underwent interventional activities such as exercise, diet, and lifestyle changes for judgment about the effectiveness of these treatments.[6] Early diagnosis and treatment of NAFLD can prevent the progression of the disease to more severe conditions such as nonalcoholic steatohepatitis (NASH), cirrhosis, and hepatocellular carcinoma.[7] Liver fat content can be evaluated by biopsy and imaging modalities such as ultrasonography (US), computed tomography, magnetic resonance imaging (MRI), and MR spectroscopy.[8]

A biopsy is a current gold standard for assessment of liver fat content; however, due to invasive nature, this method has a number of limitations such as risk of bleeding, infection, and bile leakage.[9] Moreover, this procedure is not reproducible and associated with sampling errors.[10] Hence, there have been comprehensive efforts to substitute this procedure by noninvasive imaging modalities.

US considered as a rapid and low-cost instrument for assessment of hepatic steatosis whereas this subjective modality suffered from low sensitivity and accuracy.[1112] Several studies have been reported that good correlation between fat quantification by chemical-shift Gradient echo (GRE) MRI techniques and histopathological findings.[131415]

Despite the numerous advantages of GRE techniques such as rapidity, good accuracy, and ability to perform at different scanners, the main weakness of this technique is that the fat measurement may be biased by T1 and T2* relaxation effects. Modifying the imaging parameters that control T1 and T2* weighting may lead to miscalculation of liver fat quantity. Although comprehensive research[1415161718] has been carried out on quantifying liver fat by GRE MR imaging, only a few studies have attempted to investigate the effects of the imaging parameters that control the T1 and T2* weighting on in vivo liver fat quantification.[19202122] Hence, this study was set out to examine the GRE protocols with different T1 and T2* weighting for liver fat quantification in patients with NAFLD.

Methods

Study design and patient population

This prospective, cross-sectional, single-center study was approved by the local ethical committee. In a period of >2 years (from September 2015 to November 2017), 34 adult patients over 18 years with biopsy-confirmed NAFLD in consistent with NASH Clinical Research Network (NASH CRN)[23] were recruited from Golestan Educational Hospital of Ahvaz, Iran, to participate in this research. Exclusion criteria were as follows: patients with pregnancy (none), alcohol consumption (none), use of therapeutic interventions between biopsy and MRI test (3 cases), MRI contraindications (1 case), and known history of other hepatic diseases. The mean time interval from biopsy to MRI examination was 21 days (range = 8–34 days). The purpose of the study was clearly explained and written informed consent was obtained from all participants.

Liver biopsy and histopathological assessment

Ultrasound-guided needle biopsies were performed with 18G from the right lobe of the liver. The mean specimen size was 2 cm × 0.1 cm fixed in formalin for 24 h, then 0.4 μm sections of the tissue stained with hematoxylin and eosin, trichrome, and reticulin. Two hepatic pathologists who were blinded to the MRI reports examined slides according to NASH CRN histological scoring system.[23] Steatosis was classified into four grades: fat deposition in 0%–5% of hepatocytes (Grade 0), 5%–33% (Grade 1), 34%–66% (Grade 2), and above 67% (Grade 3).

Magnetic resonance imaging sequences

All examinations were performed using 1.5 Tesla MR scanner (Essenza, Siemens Medical Systems, Erlangen Germany) equipped with four-channel torso coil. Sixteen GRE sequences with different T1 weighting contrast were performed for each patient with four echo times (TE1= 2.31, TE2= 4.86, TE3= 7.67, and TE4= 10.58 ms). In each sequence, three slices with 6-mm thickness, field of view = 380 mm, bandwidth = were performed from the liver at hilarious level. The taken images were included of three first out-of-phase images (OP1) for TE1, three first in-phase images (IP1) for TE2, three second out-of-phase images (OP2) for TE3, and three second in-phase images (IP2) for TE4. Therefore, 12 images were achieved from any sequence. In each sequence, repetition time (TR) or flip angle (α) parameter was changed and set out according to Table 1, and other scan parameters were fixed at all sequences [Table 1].

Table 1

Characteristics of repetition time (ms) and flip angle (α: Degree) at different sequences

Image analysis

All analyzes were performed by two experienced radiologists who were blinded to histopathological findings at the same workstation. To consider the effect of fat inhomogeneity, regions of interest (ROIs) were placed on three slices at 7 different regions of liver parenchyma by avoiding other structures such as diaphragm, major vessels and so on. ROIs which placed on OP1 images were copied into other images identically. The mean signal intensity of OP1, IP1, OP2, and IP2 images were measured separately for each sequence by averaging the 21 ROIs.

Calculation of fat indexes by magnetic resonance imaging signals

In this study, fat index (FI) values were measured for all 16 sequences separately with three different methods. Method 1: In this method, FIs (1–16) were calculated by considering signal intensity changes from first dual echoes (OP1 and IP1 images) according to Eq. 1 for all 16 sequences, respectively. Method 2: FIs values from 17 to 32 were calculated by providing this method. Like the first method and taking into account of the second dual echoes (OP2 and IP2 images), FIs (17–32) were achieved by Eq. 2. Method 3: In this method, signal loss due to T2* decay was corrected. Estimated FI by this method was achieved using Eq. 3.

Where SIOP1 and SIIP1 are out of phase and in phase corresponding to the mean signal intensity of first dual GREs, respectively, SIOP2 and SIIP2 and are corresponding to mean signal intensity of second dual GREs. △TE is the time difference between IP2 and IP1 (10.58 − 4.86 = 5.72 ms). In each sequence, three FIs were calculated with three different methods (Eqs. 1–3). Therefore, 48 FIs obtained for each patient normally. As stated in Eq. 1, FIs (1–16) which are related to sequences (1–16) were measured by considering signal intensity changes at first dual echoes. FIs (17–32) established using Eq. 2 from signal intensity changes at second dual echoes of each sequence. Finally, FIs (33–48) were obtained using Eq. 3 and correction of T2* decay effects for each sequence.

Statistical analysis

The calculated FIs by MRI are presented as a mean ± standard deviation. The correlation between histopathologic findings and MRI FIs was performed using Pearson coefficient. Data management and analysis were performed using SPSS 16.0 (SPSS Inc., SPSS for Windows, Chicago, USA). Significance levels were set at the 1% level.

Results

Demographic and clinical data

In this study, MRI sequences were performed for the thirty confirmed NAFLD adult patients (mean age: 41.4 ± 11.4 years). Characteristics and clinical data of these patients are given in Table 2.

Table 2

Demographic and clinical results for the thirty patients with nonalcoholic fatty liver disease

The results of calculated fat indexes

The first set of analysis indicated the impact of variable T1 and T2* weighting imaging parameters for calculation of FIs. Totally 48 FIs were calculated using three equations. The results obtained from the calculation of FIs are shown in Table 3. As can be seen in this table, the mean FIs (17–32) which calculated from Eq. 2 are very low and in some indexes have a negative value. The mean FIs (33–48) which obtained from Eq. 3 have the maximum values in comparison to the other FIs. The maximum mean FI was 23.58% (FI36) which corresponds to heavily T1 weighted pulse sequence (sequence 4 TR = 50 ms, α = 90° quantified by method 3. The minimum mean FI was −2.49% (FI29) which corresponds to the minimal T1 weighted pulse sequence (sequence 13 TR = 200 ms, α = 20° quantified by method 2 [Table 3].

Table 3

Calculated fat indexes by different magnetic resonance imaging sequences

Relationship between flip angle parameter and calculated fat indexes

Figure 1a–c illustrates the relationship between mean calculated FI and flip angles for different TR parameters. As can be seen in these figures, the mean calculated FI increases with rising of flip angle, especially at low flip angles. Mean FI was obtained by extrapolating the curves to flip angle → 0° for minimizing the effect of this parameter [Figure 1].

Figure 1

(a) Relationship between fat indexes calculated from Eq. 1, with flip angle at different repetition time. (b) Relationship between fat indexes calculated from Eq. 2, with flip angle at different repetition time. (c) Relationship between fat indexes calculated from Eq. 3, with flip angle at different repetition time

Relationship between calculated fat index and repetition time parameter

From Figure 2a and c, it can be seen the relationship between mean calculated FIs and TR parameter. These graphs show that the mean FI declines gradually with increasing TR parameter; however, as TR increases, calculated FI by method 2 [Figure 2b] decreased more intensely and irregularly [Figure 2].

Figure 2

(a) Relationship between fat indexes calculated from Eq. 1, with repetition time parameter at different flip angles. (b) Relationship between fat indexes calculated from Eq. 2, with repetition time parameter at different flip angles. (c) Relationship between fat indexes calculated from Eq. 3, with repetition time parameter at different flip angle

Correlational analysis

Pearson's correlation coefficients were used to determine the relationship between obtaining various FIs by MR imaging and histopathologic liver steatosis. There is a strong correlation between all 48 FIs calculated with MR imaging and liver biopsy (correlation coefficients range from r = 0.81 to r = 0.92). The results of the correlation analysis are shown in Table 4 and all correlations are significant at the 0.01 level. It can be seen from the data in Table 4 that the correlations determined with the use of methods 1 and 3 were slightly stronger than the correlation with the use of method 2 in fat calculation at all sequences. Correlation coefficients corresponding to calculations with method 2 in all sequences were improved with a flip angle increasing (r = 0.82 to r = 0.9). From Table 4, FIs from sequence with small flip angle calculated with methods 1 and 3 have a slightly better correlation than FIs with large flip angle (r = 0.92 in FI1 vs. r = 0.89 in FI4) or (r = 091 in FI33 vs. r = 0.88 in FI36) [Table 4].

Table 4

Correlation between calculated fat indexes and liver biopsy

Discussion

The present study was designed to investigate different chemical-shift GRE methods for fat quantifying in patients with NAFLD. Another more important question was to determine the effect of imaging parameters that control T1 and T2* relaxation in these methods. Our results demonstrate that first dual GRE techniques (method 1) and GRE techniques with correction of T2* decay effects (method 3) are better than second GRE techniques (method 2). Moreover, there was no significant difference observed between diagnostic accuracy of methods 1 and 3 in fat quantification. On the question of imaging parameters which affects fat measurement, this study found that flip angle parameter could be a major factor, if not the only one, causing the overestimation of liver fat content.

By referring to the data in Table 3, maximum and minimum quantified FIs are related to the sequences with maximum and minimum T1 weighting. From the literatures, phase interference is not only reason for using of GRE for fat quantification. Due to the shorter T1 relaxation, signal of fat is higher at T1 weighted sequences and leads to fat overestimation.[24] These results are consistent with those of other studies and many previous authors suggest that to mitigate T1 confounding effects, GRE sequences should be conducted with long TR and low flip angles.[262728]

As shown in Table 3, most measured FIs with method 2, especially measured FIs in patients with low fat grades (FIs: 17–32), were negative. Signal loss due to T2* decay effects leads to more reduction of signal intensity at latter IP2 images relative to earlier OP2 images. As a consequence, FIs which calculated by method 2 have negative values. Furthermore, increasing of TE and T2* decay at last dual echoes causes that signal-to-noise ratio (SNR) diminished considerably. Thus, OP2 and IP2 images typically do not have an acceptable quality. Cassidy et al.[24] recommended the use of first dual echoes to avoid this problem in practice.

Bydder et al.[22] first explained the effect of flip angle for fat measurement on three NAFLD patients. Hansen et al. experimentally illustrated direct relationship between estimated fat and flip angle.[20] Our findings are consistent with their studies and suggest that modifying flip angle affects T1 weighting clearly and as a consequence measured FI.[22] Theoretically, another parameter that controls T1 weighting is TR. Increasing of TR parameter (slight decrease of T1 weighting) decreases is calculated fat degree gradually. However, this parameter is not as effective as flip angle. Hansen et al.[20] findings show that the effect of TR on fat measurement did not statistically significant. In their study, TRs above 130 ms were examined. According to the results of this study, the amount of FIs remains constant at TRs above 100 ms. The data reported here appear to support the assumption that the TRs above 100 ms do not have a significant effect on T1 weighting contrast and determination of fat content.

Our findings show that the correlation coefficient of calculated FIs by methods 1 and 3 was more reliable than calculation by method 2. Weaker SNR at last echoes relative to first echoes could be the major reason. As in this method, correlation coefficients improve when flip angle increases (raising SNR). In accordance with the present study, some previous studies demonstrate that the correlation between calculated fat and biopsy findings was same with and without correction of T2* decay effects.[1729] Conversely our result is not in agreement with reported data by Westphalen et al that indicated the liver iron as a potential pitfall for fat quantification.[30] This variation could be due to differences in the sample characteristics. Evidence from this study suggests that, in fat quantification with GRE techniques, it is possible that the third and fourth echoes are unnecessary. Furthermore, these last echoes subject to noise and maybe prolonged scan time and limit anatomy coverage.

Due to the nature of cross-sectional study, current investigation was limited by small sample size and caution must be applied, as the findings might not be transferable to NAFLD patients with fibrosis or iron deposition. Thus, more research will need to be done to determine the fat content in patients with these conditions.

Conclusion

T1 relaxation effects probably more critical than T2* to measurement of fat content. Our findings suggest that flip angle parameter could be a major factor, if not the only one, causing the overestimation of liver fat content. The sequences with low flip angle are more suitable for fat quantification. In this case, using of long TR is recommended to maintain SNR and image quality. However, increasing of scan time should be considered. The results of this research can be useful for adjustment of imaging parameter in GRE techniques to measurement of fat content.

Financial support and sponsorship

This study is funded by Vice-Chancellor for Research Affairs of Ahvaz Jundishapur University of Medical Sciences (Grant number: RDC-9509).

Conflicts of interest

There are no conflicts of interest.

BIOGRAPHIES

Mansour Zabihzadeh received his B.Sc degree in physics in 2000 from Shahid Rajaee University of Karaj, Karaj, Iran; and he obtained his Master and PhD degree in medical physics from Tehran University of medical Sciences, Tehran, Iran in 2003 and 2009 respectively, he is now an associated professor at Ahvaz Jundishapur University of medical Sciences. His research interest include Radiotherapy, radiation dosimetry, Image analysis.

Email: manzabih@gmail.com

Mohammad Momen Gharibvand received his MD degree from Shiraz University of medical Sciences, Shiraz, Iran, in 1999 and his radiology Specialty from Shahid Beheshti University of medical Sciences, Tehran, Iran in 2005, he is currently assistant professor at Ahvaz Jundishapur University of medical Sciences, His research interests include intervention radiology, image analysis, quantitative imaging.

Email: mohamad.momen2017@gmail.com

Azim Motamedfar received his MD degree from Tabriz University of medical Sciences, Tabriz, Iran, in 2001 and his radiology Specialty from Tehran University of medical Sciences, Tehran, Iran in 2006, he is currently assistant professor at Ahvaz Jundishapur University of medical Sciences, His research interests include intervention radiology, image analysis, quantitative imaging.

Email: AZEXIOUS@yahoo.com

Morteza Tahmasebi received his MD and radiology Specialty degree from Tehran University of medical Sciences, Shiraz, Iran, in 1989 and 2005, respectively, he is currently associated professor at Ahvaz Jundishapur University of medical Sciences, His research interests include intervention radiology, dynamic imaging, quantitative imaging.

Email: tahmasebilmn@yahoo.com

Amir Hossein Sina received his MD degree from Mashhad University of medical Sciences, Mashhad, Iran, in 2004 and his pathlogy Specialty from Iran University of medical Sciences, Tehran, Iran in 2010, he is currently assistant professor at Ahvaz Jundishapur University of medical Sciences, His research interests in the field of pathlogy disease assessment.

Email: dramirhoseinsina@yahoo.com

Kavous Bahrami received his B.Sc degree in radiology technology from Ahvaz Jundishapur University of medical Sciences, Ahvaz, Iran, in 1998, He is currently working at the MRI department of Golestan hospital, Ahvaz, Iran. his research interests are in the fields of imaging technology.

Email: kavous_bahrami@yahoo.com

Mozafar Naserpour received his B.Sc degree in radiology technology from Tehran University of medical Sciences, Tehran, Iran in 2005, and he got his M.Sc and PhD degrees from Ahvaz Jundishapur University of medical Sciences in 2011 and 2018 respectively, he is now an assistant professor at Behbahan Faculty of medical Sciences, his research interests are in the fields of imaging technology, radiation dosimetry, signal processing. Quantitative MRI.

Email: naserpour1983@gmail.com