Diagnostic performance comparison between ultrasound attenuation measurements from right and left hepatic lobes for steatosis detection in non-alcoholic fatty liver disease

Abstract

Rationale and Objectives:

Non-alcoholic fatty liver disease (NAFLD) is currently diagnosed by liver biopsy or MRI proton density fat fraction (MRI-PDFF) from left hepatic lobe (LTHL) and/or right hepatic lobe (RTHL). The objective of this study was to compare the diagnostic value of ultrasound attenuation coefficients (ACs) from RTHL and LTHL in detecting hepatic steatosis using biopsy or MRI-PDFF as a reference standard.

Materials and Methods:

Sixty-six patients with suspected NAFLD were imaged with an Aplio i800 ultrasound scanner (Canon Medical Systems, Tustin, CA). Five AC measurements from RTHL and LTHL were averaged separately and together to be compared with the reference standard.

Results:

Forty-seven patients (71%) were diagnosed with NAFLD. Mean ACs were significantly higher in fatty livers than non-fatty livers (RTHL: 0.73 ± 0.10 vs. 0.63 ± 0.07 dB/cm/MHZ; p<0.0001, LTHL: 0.78 ± 0.11 vs. 0.63 ± 0.06 dB/cm/MHz; p<0.0001, RTHL & LTHL: 0.76 ± 0.09 vs. 0.63 ± 0.05 dB/cm/MHz; p<0.0001). Biopsy steatosis grades (n=31) were better correlated with the mean ACs of RTHL & LTHL (r=0.72) compared to LTHL (r=0.67) or RTHL (r=0.61). Correlation between MRI-PDFF (n=35) and mean ACs was better for LTHL (r=0.69) compared to the RTHL & LTHL (r=0.66) or RTHL (r=0.45). Higher diagnostic accuracy was shown for the mean ACs of RTHL & LTHL (AUC 0.89, specificity 94%, sensitivity 78%) compared to LTHL (AUC 0.89, specificity 88%, sensitivity 82%) or RTHL (AUC 0.81, specificity 89%, sensitivity 68%).

Conclusion:

Ultrasound ACs from RTHL and LTHL showed comparable diagnostic values in detection of hepatic steatosis with the highest diagnostic accuracy when they were averaged together.

INTRODUCTION

Nonalcoholic fatty liver disease (NAFLD) is a condition with the accumulation of more than 5% fat in the liver cells (steatosis) in the absence of excess alcohol intake or other chronic liver diseases (chronic viral hepatitis, autoimmune and other metabolic liver diseases, and the use of medications that can induce steatosis) (1, 2). It encompasses a spectrum of liver disorders ranging from hepatic steatosis to nonalcoholic steatohepatitis (NASH) and may progress to cirrhosis and hepatocellular carcinoma with a higher risk of liver-related death (3–5). NAFLD is predicted to become the most common indication for liver transplants in the near future (2).

Currently, liver biopsy is considered the gold standard for the diagnosis of NAFLD. However, liver biopsy has limitations such as sampling errors due to small tissue samples obtained from either RTHL or LTHL, invasiveness, operator dependency, and bleeding complications. Considering the potential risks of the procedure, it is not an optimal diagnostic measure for all NAFLD suspects especially those at their early stages or requiring multiple consecutive assessments (6).

A magnetic resonance imaging technique that measures the proton density fat fraction (MRI-PDFF) is also accepted as a clinical reference standard for tissue fat concentration and is well correlated with histologic steatosis grade by liver biopsy (7, 8). In this MRI-based quantitative fat measurement technique, a single PDFF estimate for the MRI exam is driven based on averaged PDFF values obtained from multiple liver segments (9). A significant difference in fat content and variability between RTHL and LTHL and liver segments has been shown among adults with NAFLD (10). This technique also has its own drawbacks such as high cost and several contraindications, such as claustrophobia and metallic implants.

Hence, there is an urgent need for a non-invasive, affordable, easily accessible, and reproducible imaging technique for the clinical assessment of hepatic steatosis in NAFLD. Emerging quantitative ultrasound-based imaging methods have shown promise in fulfilling these requirements. One of the well-studied ultrasound parameters is attenuation coefficient (AC) which has shown excellent results compared to the clinical reference standards of liver biopsy and MRI-PDFF in estimating the severity of hepatic steatosis (11). A frequency-dependent attenuation happens during the propagation of the ultrasound beam through liver tissue. The AC is calculated in decibel per centimeter per megahertz (dB/cm/MHz). The estimates of ACs have been shown to be higher among fatty versus non-fatty livers in previous studies (12, 13).

According to a meta-analysis, Jang et al. have shown a good overall diagnostic performance in detecting hepatic steatosis using AC (13). Currently, in almost all the reported ultrasound studies that investigated the AC for diagnosis of hepatic steatosis, scanning has been performed only in the RTHL through the intercostal approach (14–26). However, RTHL and/or LTHL are both considered for liver biopsy or MRI-PDFF.

In this study, we aimed to investigate if the ACs from LTHL can be useful in detecting NAFLD independently and in combination with the ACs from RTHL. These measurements were compared with a reference standard of liver biopsy or MRI-PDFF.

MATERIALS AND METHODS

Study subjects

This study was approved by our institutional review board and subjects signed informed consent. Between December 2020 and May 2022, we prospectively enrolled a total of 79 patients who met the eligibility criteria as follows: subjects 18 years and older, known or suspected NAFLD scheduled for a clinical standard of care liver biopsy or hepatic MRI within 1 month, or known NAFLD or non-fatty liver proven by a liver biopsy or hepatic MRI results performed within 1 month prior. Patients were excluded in presence of significant alcohol consumption (>21 drinks/week in men and >14 drinks/week in women), alternate diagnosis of chronic liver disease (i.e., chronic viral hepatitis, metabolic, autoimmune, or drug-induced liver disease), a history or presence of liver malignancy (any size), or focal liver lesion > 5 cm.

2D ultrasound attenuation imaging (ATI)

A diagnostic ultrasound scanner (Aplio i800, Canon Medical Systems, Tustin, CA) with built-in ATI software was utilized to obtain the ACs. All subjects were imaged in a supine position with a convex transducer (i8CX1, frequency bandwidth of 1–8 MHZ). The transducer was located perpendicular to the liver capsule and angled to avoid liver vascularity and rib shadowing. Images from the RTHL and LTHL were obtained from the intercostal and subxiphoid windows, respectively.

With the guide of simultaneous B mode ultrasound, a sample box (where ACs are obtained) was positioned within homogeneous parenchyma in the RTHL or LTHL. Figure 1 demonstrates an example of ATI technology using the color-coded map in the RTHL and LTHL. A color-coded map based on ACs was visualized inside the sample box with vacancies referring to filtered large vessels or areas with strong artifacts (reverberation) (16). Subjects were instructed to hold their breath for a few seconds to reduce the motion artifacts (which may cause vacancies in a color-coded AC map) during the image acquisition. The size of the sample box was adjusted according to the size of the liver lobes and placed in areas of homogenous liver parenchyma to obtain a homogeneous AC color map with the maximum signal. Once the size and location of the sample box were adjusted, 2–3 seconds of ATI clip were obtained.

Figure 1.

Examples of AC acquisition in a 55-year-old female diagnosed with fatty liver (a and b) and a 74-year-old male with non-fatty liver (c and d) using ATI (AC: attenuation coefficient, ATI: attenuation imaging, LTHL: left hepatic lobe, RTHL: right hepatic lobe).

From the 2–3 second ATI clip, frames were acquired and, on each frame, a trapezoid shape region of interest (ROI) with a consistent size of 3 cm height and width was placed within the sample box. A 3 cm ROI size was the smallest possible by the ATI software and was determined to avoid possible artifacts, especially in the LTHL which normally has a smaller scanning area compared to the RTHL. The ROI was placed at least 2 cm below the liver capsule to avoid dark orange areas at the top of the sample box (representing the reverberation artifact of the liver capsule) as well as any dark blue areas at the bottom of the sample box (representing weakened signals) (16, 22, 27). Additionally, the ROI was positioned not to include the blank area (representing no AC available) more than 50% of the whole ROI area. The average of ACs within the ROI was automatically displayed on the screen along with the R² value (indicating the goodness of fit to obtain ACs). The R² value was utilized to confirm the optimal ROI placement and accuracy. ACs were considered valid when R² ≥ 0.80, good if (R² = 0.80–0.89), and excellent if R² ≥ 0.90. Five ACs (from the ROIs on 5 ATI frames) with R² ≥ 0.90 and interquartile range (IQR)/median ≤0.30 were selected and averaged to obtain the final AC of each liver lobe (16).

MRI-PDFF

The MRI- PDFF values were quantified with 4 measurements from the 2 ROIs each in RTHL and LTHL by a radiologist. The absolute difference between MRI-PDFF values in RTHL and LTHL was calculated to reflect the variability of fat content between two liver lobes. Liver steatosis was defined by an MRI- PDFF of greater than 6.4%. Mild, moderate, and severe steatosis were defined by PDFF of 6.5–17.4%, 17.5–22.1%, and 22.2% or greater respectively, following our institutional clinical guideline (28).

Liver biopsy

Subjects’ biopsy results were also obtained through the electronic health record system (Epic Systems Corporation, Verona, WI). Liver steatosis was defined as the presence of fat in greater than 5% of the hepatocytes. Mild, moderate, and severe steatosis were defined as 5–33%, 34–66%, and 67% or greater, respectively.

Statistical analyses

Student t-test was performed to compare ACs between fatty and non-fatty livers. AC variability between two liver lobes was assessed by the mean absolute difference between AC values in RTHL and LTHL. The correlations between ACs and MRI- PDFF values were analyzed with the Pearson rank correlation coefficient. The correlations between biopsy steatosis grades with ACs were analyzed by Spearman’s rank correlation test. The correlation was categorized as follows based on the correlation coefficient value: 0.00–0.25, none or slight; 0.25–0.50, fair to moderate; 0.50–0.75, moderate to good; and 0.75– 1.00, almost perfect (29). The diagnostic performance of ATI was assessed with receiver operating characteristic (ROC) analysis. The optimal cutoff point was determined where the Youden index was maximum (30). The area under the ROC curve (AUC) of ATI for hepatic steatosis detection was assessed by using 1, 2, 3, 4, and 5 AC measurements of the RTHL, LTHL, and average of RTHL&LTHL. All statistical analysis was performed using GraphPad Prism 9 (GraphPad Software, San Diego, California, USA) with a significance level of 0.05.

RESULTS

Seventy-nine subjects were enrolled; 10 subjects were excluded due to pathologic results (n=8) (such as drug-induced steatosis, autoimmune disease, or viral hepatitis) or unavailability of their MRI-PDFF (n=2). While the ACs from an RTHL could be obtained for all enrolled subjects, the ACs from the LTHL were not available for 3 subjects (all with non-fatty liver) due to the small ATI area resulting in unreliable ACs; it was not possible to avoid reverberation artifacts or fit the trapezoid ROI box inside the LTHL (c.f. figure 2). Thus, 66 subjects’ data were included in the analyses. Among 66 subjects, 31 underwent liver biopsy (21 from the RTHL and 10 from the LTHL) and 35 had an MRI-PDFF. In the biopsy group, 26 out of 31 were diagnosed with liver steatosis. In the MRI-PDFF group, 21 out of 35 were diagnosed with liver steatosis. Table 1 is a summary of the demographic characteristics of the study cohort (Table 1).

Figure 2.

An example of AC acquisition limitation in LTHL due to the small size of the LTHL, especially in thin patients (a and b were acquired from patients with BMI of 18 and 20 kg/m², respectively). The artifacts could not be avoided during the placement of the sample box and/or ROI box in the LTHL (AC: attenuation coefficient, BMI: body mass index, LTHL: left hepatic lobe).

Table 1.

Demographic information of the study subjects (BMI: body mass index; std: standard deviation).

Variable	Subjects with fatty liver (n=47)	Subjects with non-fatty liver (n=19)
Age (year) (mean ± std)
• Female	55 ± 12.5 (n=30)	46 ± 16.3 (n=9)
• Male	56 ± 12.4 (n=17)	53 ± 19.0 (n=10)
BMI (kg/m2) (mean ± std)
• Female	34 ± 6.7 (n=30)	30 ± 7.2 (n=9)
• Male	32 ± 6.5 (n=17)	30 ± 6.4 (n=10)
Ethnicity
White or Caucasian	38 (70%)	14 (74%)
Black or African American	5 (11%)	4(21%)
Asian	3 (6%)	1 (5%)
Hispanic	1 (2%)	0

In the biopsy group, steatosis grades varied from 0 to 3, and the mean value ± standard deviation (std) was 1.42 ± 0.83. The mean biopsy steatosis grade was higher in the patients with biopsies in the LTHL than in the RTHL (1.50 ± 0.85 vs. 1.38 ± 0.87). For this group, ACs were significantly higher in LTHL compared to RTHL among patients with fatty liver (0.78 ± 0.11 vs. 0.72 ± 0.10 dB/cm/MHz; p=0.04), but not in non-fatty livers (0.63 ± 0.06 vs. 0.62 ± 0.08 dB/cm/MHz; p=0.81).

The mean MRI-PDFF values (average of RTHL & LTHL) varied from 2% to 39%, and the mean value ± std was 12.03 ± 9.26%. MRI-PDFF values were higher in RTHL compared to LTHL, though they were not statistically significant (mean ± std: 12.60 ±9.49 vs. 11.47± 9.15 %, respectively; p=0.61). There was no significant difference between RTHL and LTHL in terms of MRI-PDFF values among patients with fatty livers (17.81 ± 8.89 vs. 16.35 ± 8.88 %, respectively, p=0.34) or non-fatty livers (4.78 ± 2.03 vs. 4.15 ± 1.31 %, respectively; p=0.60). Similarly, AC values in the MRI group were not significantly different between RTHL and LTHL in fatty livers (0.74 ± 0.10 vs. 0.77 ± 0.11 dB/cm/MHz, respectively; p=0.36) and non-fatty livers (0.65 ± 0.06 vs. 0.64 ± 0.06 dB/cm/MHz, respectively; p=0.77).

Variability between MRI-PDFF values in RTHL and LTHL was calculated for different grades of steatosis; grade 0 (n=14), grade 1 (n=11), grade 2 (n=5), and grade 3 (n=5) had mean variability of 1.39%, 1.46%, 3.43%, and 1.41%, respectively. Variability in ACs between RTHL and LTHL in the same group of patients were 0.077, 0.103, 0.099, and 0.078 dB/cm/MHz, respectively.

The ACs in the RTHL were significantly higher in fatty livers than in non-fatty livers (mean ± std: 0.73 ± 0.10 vs 0.63 ± 0.07dB/cm/MHZ, respectively) (p < 0.001). Additionally, the ACs in the LTHL and averages of the RTHL & LTHL ACs showed a significant difference between fatty and non-fatty livers (0.78 ± 0.11 vs. 0.63 ± 0.06 dB/cm/MHZ; p < 0.0001 and 0.76 ± 0.09 vs. 0.63 ± 0.05 dB/cm/MHZ; p <0.0001, respectively). Example ATI of fatty and non-fatty livers are shown in figure1. The results of ROC curve analysis of ATI in identifying fatty livers are presented in figure 3 and Table 2.

Figure 3.

ROC curves of the RTHL vs. LTHL vs. RTHL & LTHL (LTHL: left hepatic lobe, RTHL: right hepatic lobe).

Table 2.

Results from ROC curve analysis (AC: attenuation coefficient, ROC: receiver operating characteristic, LTHL: left hepatic lobe, RTHL: right hepatic lobe).

	Mean AC of RTHL	Mean AC of LTHL	Mean AC of LTHL & RTHL ACs
Area under the ROC curve	0.81	0.89	0.89
P value	<0.0001	<0.0001	<0.0001
Cutoff value	> 0.7120	> 0.6830	> 0.6840
Specificity at the cutoff value	68%	83%	79%
Sensitivity at the cutoff value	89%	89%	95%

In the biopsy group, the RTHL AC, LTHL AC, and an average of RTHL & LTHL ACs showed a good correlation with the steatosis grades (r=0.61, r=0.67, and r=0.72, respectively). When matching the biopsy location and AC measurement location, the correlation (r) was improved in the LTHL (n=10; r= 0.88) but not in the RTHL (n=21; r=0.55). The mean ACs from RTHL (n=21) and LTHL (n=10) were 0.70 and 0.79 dB/MHz/cm, respectively and showed a linear relationship with the mean biopsy grades from RTHL (1.38) and LTHL (1.50).

The mean MRI-PDFF values (average of RTHL & LTHL values) showed a good correlation with the LTHL ACs and an average of RTHL & LTHL ACs (r=0.69 and r=0.66, respectively). The correlation between the mean MRI-PDFF values and RTHL ACs was weaker (r=0.45). The correlation between MRI-PDFF values of RTHL and LTHL with the matching lobe ACs was r=0.43 and r=0.67, respectively.

An analysis was performed to assess the diagnostic accuracy variability of ATI for hepatic steatosis detection by the number of AC measurements from RTHL and LTHL. Using two measurements each in the RTHL and LTHL, AUC for RTHL, LTHL, and the average of RTHL & LTHL was 0.80, 0.88, and 0.88, respectively, compared to the AUC of 5 measurements inside the RTHL as 0.81 (cf., Table 3).

Table 3.

Diagnostic accuracy of ATI for hepatic steatosis detection by various numbers of AC measurements in RTHL, LTHL, and an average of RTHL&LTHL.

	Number of AC measurements in each liver lobe	1	2	3	4	5
RTHL	AUC	0.75	0.80	0.81	0.81	0.81
	P value	0.0015	0.0002	<0.0001	0.0001	<0.0001
LTHL	AUC	0.87	0.88	0.88	0.89	0.89
	P value	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
RTHL&LTHL	AUC	0.86	0.88	0.88	0.88	0.89
	P value	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001

DISCUSSION

While liver biopsy and MRI-PDFF are clinical reference standards for the diagnosis of hepatic steatosis, it is believed that ultrasound-based liver fat quantification has the potential to be implemented into daily practice (10, 13). During a liver biopsy, specimens might be collected from either RTHL or LTHL, and in MRI-PDFF several ROIs are picked inside both RTHL and LTHL. However, most ultrasound studies to date have only used images from the RTHL for liver fat quantification, following the liver stiffness imaging protocol recommended by the Society of Radiologists in Ultrasound to maintain the acquisition requirement for accurate simultaneous liver stiffness values (27). Acknowledging the possible uneven distribution of the fat in the liver and bi-lobar coverage of the liver in both clinical reference methods, a comparable ultrasound-based fat quantification is needed, therefore in this study, ACs of the LTHL have been evaluated if it may add diagnostic value to the current RTHL only ultrasound protocol (31).

In our study, the LTHL ACs showed a good diagnostic performance for detecting hepatic steatosis. This demonstrates that we may use the LTHL ACs to diagnose fatty liver in case of any occupying mass that limits ultrasound scanning inside the RTHL. However, there were some limitations of LTHL image acquisition using ATI, such as small scanning area, especially in subjects with normal or low BMIs. Even using the smallest available ROI box of the built-in ATI software, reliable ACs could not be obtained in 3 cases. This may be resolved by modifying the software to allow a smaller ROI box in the future, which can also help with the pediatric application. Mean ACs were significantly higher in the LTHL compared to the RTHL in the fatty liver group, but this difference was not significant in the non-fatty liver group. We also observed a higher mean steatosis grade in LTHL biopsy results compared to the right, well correlated with higher ACs in the corresponding LTHL. The overall higher attenuation values in LTHL may be attributed to higher LTHL fat accumulation in the biopsy group. However, our limited clinical data (actual fat content was available only from one side of the liver lobe by biopsy) did not allow us to compare the actual fat contents in both liver lobes of the same patient.

Capitan et al., found a significant difference in terms of the degree of steatosis (by triple-echo MR technique) between the RTHL and LTHL among both fatty (n=59) and non-fatty (n=62) livers in patients with type-2 diabetes, with a higher fat content inside the RTHL (32). There are also other reports about uneven fat accumulation in fatty livers, which found significantly higher fat content in the RTHL compared to the LTHL by MRI in both non-diabetic and NAFLD patients (33, 34). Alternatively, in some studies at a microscopic level, when liver biopsy specimens were taken from the RTHL and LTHL, minimal variability and excellent agreement were demonstrated in terms of steatosis among NAFLD-suspected patients (35–38). In this study, no significant difference was found in MRI-PDFF values and/or ACs between RTHL and LTHL among NAFLD patients. The study by Lee et al. found the least intrahepatic variability in terms of the degree of steatosis by MRI-PDFF among obese patients (BMI≥25) with grade 0 fatty liver (fat fraction: 0–6.5%) (34), which is consistent with our results from the MRI group. Both ACs and MRI-PDFF values in the RTHL and LTHL showed the least variability in the patients with grade 0 of steatosis.

While the ACs of the RTHL showed a good correlation with steatosis grading by liver biopsy (r=0.68), it was weak with MRI-PDFF (r=0.45). This could be due to a case in the MRI group, with the highest AC variability between RTHL and LTHL among the study population. The ACs from RTHL and LTHL were 0.48 dB/MHz/cm and 0.73 dB/MHz/cm, respectively for this case whose mean MRI-PDFF was 21%. While the AC from RTHL is lower than the widely known normal liver AC of 0.5 dB/MHz/cm, the AC from LTHL indicates the possibility of fatty liver. Therefore, this can be an example of how measuring ACs from both RTHL and LTHL can arouse suspicion and possibly help with the diagnosis of liver steatosis. Additionally, this case may have caused a low correlation between the MRI-PDFF and RTHL ACs. Excluding this case significantly improved the correlation between MRI-PDFF and RTHL AC from r=0.45 to r=0.58. Regarding the correlation between MRI-PDFF and AC in the RTHL, Ferraioli et al. (19) showed an almost perfect correlation (r = 0.78), and a good correlation was observed by Kwon et al. (r = 0.75) (24). However, in the study by Ferraioli et al., the median (IQR) range of MRI-PDFF detected among all included subjects was 6.2 (3.9–14) % compared to 8.4 (4.8–18.6) % in our study. We included patients who were highly suspicious or known cases of NAFLD, which may justify the higher MRI-PDFF reported in our study.. Kwon et al. study population was different from ours, as they included subjects with different liver diseases, including cholecystitis and viral hepatitis, with 53% of them having MRI-PDFF <5.1 % which explains the better correlation between ATI and MRI-PDFF compared to our results. In our study, mean RTHL & LTHL ACs showed a better correlation with MRI-PDFF compared to the RTHL ACs among patients with different stages of steatosis (r=0.66 vs. r=0.45).

The averaged ACs obtained from the RTHL & LTHL showed good diagnostic value in our study. Similarly, one study has shown that overall histopathologic diagnostic accuracy for NAFLD was improved by combining biopsy samples of the RTHL and LTHL (37). Additionally, previous studies of MRI-PDFF have shown that ROI sampling of both RTHL and LTHL provided a close estimate of the liver mean PDFF (39–41). This study also showed that an average of ACs for RTHL&LTHL with only one measurement inside each liver lobe has a better diagnostic performance compared with the average of five measurements inside the RTHL (AUC 0.86 vs. 0.81).

This study has limitations, such as the small number of participants and a smaller number of non-fatty livers compared to fatty livers due to inclusion criteria.

CONCLUSION

ACs acquired from either the RTHL or LTHL showed a high potential to detect hepatic steatosis. Moreover, the diagnostic accuracy was improved by averaging ACs from the RTHL and LTHL.

Abbreviation:

AC

attenuation coefficient

ATI

attenuation imaging

AUC

area under the curve

BMI

body mass index

LTHL

left hepatic lobe

MRI-PDFF

MRI-proton density fat fraction

NAFLD

non-alcoholic fatty liver disease

ROC

receiver operating characteristic

ROI

region of interest

RTHL

right hepatic lobe

std

standard deviation