Quantitative Evaluation of Hepatic Steatosis Using Advanced Imaging Techniques: Focusing on New Quantitative Ultrasound Techniques

Abstract

Nonalcoholic fatty liver disease, characterized by excessive accumulation of fat in the liver, is the most common chronic liver disease worldwide. The current standard for the detection of hepatic steatosis is liver biopsy; however, it is limited by invasiveness and sampling errors. Accordingly, MR spectroscopy and proton density fat fraction obtained with MRI have been accepted as non-invasive modalities for quantifying hepatic steatosis. Recently, various quantitative ultrasonography techniques have been developed and validated for the quantification of hepatic steatosis. These techniques measure various acoustic parameters, including attenuation coefficient, backscatter coefficient and speckle statistics, speed of sound, and shear wave elastography metrics. In this article, we introduce several representative quantitative ultrasonography techniques and their diagnostic value for the detection of hepatic steatosis.

INTRODUCTION

Nonalcoholic fatty liver disease (NAFLD) is a spectrum of liver abnormalities that present with excessive fat accumulation [1,2,3]. The prevalence of NAFLD has been steadily increasing, and it is currently the most common chronic liver disease in Western countries as well as in Asia [4,5,6,7,8]. Although simple fatty liver, also called nonalcoholic fatty liver (NAFL), is generally considered non-progressive, it can progress to nonalcoholic steatohepatitis (NASH) and clinically significant liver fibrosis [9]. In addition, the increased degree of hepatic steatosis in NAFLD is associated with a higher prevalence of metabolic syndrome and increased cardiovascular risk [10]. Therefore, efforts are actively being made to treat NAFLD. Hepatic steatosis in NAFLD seems to be reversible through treatment, including lifestyle interventions [11]. Furthermore, hepatic steatosis is frequently found in other chronic liver diseases, such as chronic hepatitis C, and the degree of hepatic steatosis is possibly associated with the hepatic fibrosis progression rate in a specific genotype of chronic hepatitis C [12]. Therefore, the detection and grading of hepatic steatosis are important for prognostication and management decisions for patients with NAFLD and other chronic liver diseases.

Currently, liver biopsy is considered the gold standard for the diagnosis and severity assessment of hepatic steatosis [3,13]. However, liver biopsy has intrinsic limitations of sampling errors and its invasiveness hinders its use. Accordingly, there is a need for reliable non-invasive biomarkers for the assessment of hepatic steatosis [14]. At present, MR spectroscopy (MRS) and MRI-proton density fat fraction (MRI-PDFF) have been accepted as non-invasive reference standards for quantifying hepatic steatosis [15,16,17]. However, these MR-based techniques have limitations, as they are expensive and not readily accessible. In contrast, conventional B-mode ultrasound is inexpensive and easily accessible, and it has been widely used for the assessment of hepatic steatosis in clinical settings despite its subjectivity [18]. Recently, various quantitative ultrasound (QUS) techniques that quantitatively characterize tissue microstructure using inherent ultrasound tissue properties, have been developed and actively validated for the diagnosis of hepatic steatosis [19,20,21,22,23,24,25].

Here, we will briefly review conventional imaging techniques for hepatic fat quantification, and discuss the basic concepts and recent advances in QUS techniques and their diagnostic performance in hepatic fat quantification. In addition, the unmet needs of the current QUS techniques and the future direction of development for the evaluation of NASH/NAFLD will be briefly discussed.

Conventional Imaging Techniques for Liver Fat Quantification

B-Mode Ultrasound

B-mode ultrasound is the most common imaging modality used to evaluate hepatic steatosis. Using B-mode ultrasound, hepatic steatosis can be graded based on the following findings: 1) higher echogenicity of the liver than that of the renal cortex, 2) impaired visualization of the intrahepatic vessels, and 3) impaired visualization of the diaphragm and posterior right hepatic lobe due to ultrasound beam attenuation (Fig. 1) [26]. Although B-mode ultrasound has the advantages of high accessibility and low cost, especially compared with MRI, it is limited by its relatively low sensitivity for detecting mild hepatic steatosis (73.3% for detection of > 0%–5% steatosis) [27] and its substantial intra- and inter-observer variability (κ = 0.54 and 0.43, respectively) [28].

Fig. 1. Conventional ultrasound images of patients with different degrees of hepatic steatosis in subcostal view (left), intercostal view (middle), and longitudinal view with right kidney (right).

A. A 37-year-old female patient without hepatic steatosis. Echogenicity of the liver (L) is similar to that of the right kidney (K). Hepatic veins (arrowheads), wall of portal veins (thin arrows), and diaphragm (thick arrows) are all clearly visualized. B. A 20-year-old female patient with mild hepatic steatosis. Echogenicity of the liver (L) is higher than that of the right kidney (K). However, hepatic veins (arrowheads), wall of portal veins (thin arrows), and diaphragm (thick arrows) are all visualized. C. A 60-year-old female patient with moderate hepatic steatosis. Echogenicity of the liver (L) is higher than that of the right kidney (K). Hepatic veins (arrowheads) and wall of portal veins (thin arrows) are partly blurred due to ultrasound beam attenuation, but the diaphragm (thick arrows) is still visualized. D. A 49-year-old male patient with severe hepatic steatosis. Echogenicity of the liver (L) is markedly higher than that of the right kidney (K). The wall of the portal vein (thin arrows), as well as the diaphragm (thick arrows), are blurred due to ultrasound beam attenuation.

CT

On unenhanced CT images, the normal liver parenchyma has a slightly higher attenuation than the spleen, whereas the hepatic attenuation value (Hounsfield unit [HU]) decreases with increasing severity of hepatic steatosis [29,30]. The generally accepted criteria for diagnosis of hepatic steatosis (hepatic fat content ≥ 30%) on unenhanced CT are as follows: 1) the absolute attenuation of the liver is less than 40 HU [31], or 2) the attenuation of the liver is at least 10 HU less than that of the spleen [32]. CT can detect hepatic steatosis with high specificity (93.5%, 88.1%, and 94.6% for diagnosis of > 0%–5%, > 10%–20% and > 25%–33% steatosis, respectively) but has a relatively low sensitivity, especially for mild cases (46.1%, 57.0%, and 72.0% for detection of > 0%–5%, > 10%–20% and > 25%–33% steatosis, respectively) [27]. Furthermore, hepatic attenuation can also be affected by other factors, including iron or glycogen deposition and drug therapy (e.g., amiodarone), which can act as confounders [33]. More importantly, exposure to ionizing radiation discourages its widespread use for the diagnosis of hepatic steatosis.

MR-Based Methods

MR-based techniques have been extensively validated as quantitative tools for hepatic steatosis [27,34,35]. At present, PDFF measured by these MR-based techniques is accepted as a noninvasive reference standard for hepatic steatosis, and may replace liver biopsy [15,16,17]. There are two major MR-based techniques for hepatic fat quantification: MRS and multi-echo Dixon MRI.

MR Spectroscopy (MRS)

¹H-MRS is based on the difference between the precession frequencies of protons in different chemical moieties (chemical shifts) [27,36,37]. To obtain MRS-PDFF of the liver, a localization voxel with dimensions of 2 × 2 × 2 cm³ or 3 × 3 × 3 cm³ is typically placed within the right hepatic lobe to avoid the large intrahepatic vessels and liver edge (Fig. 2A) [38]. Then, the PDFF of a target volume can be calculated by adding all the individual lipid peak areas in the MRS frequency spectrum and dividing it by the sum of the lipid and water peaks (Fig. 2B) [33,36].

Fig. 2. MRS (A, B) and MRI-proton density fat fraction (C) results of a patient with hepatic steatosis.

A. A 3 × 3 × 3 cm-sized voxel was placed in the right hepatic lobe for MRS. B. The MRS frequency spectrum (upper) at the voxel shows a large water peak at 4.7 ppm as well as multiple lipid peak areas at a lower ppm. Fat and water signals at different TEs are plotted (lower), and the T2-corrected fat signal fraction is calculated by 29.3%. C. Fat signal fraction map acquired with multi-echo Dixon MRI in the same patient: brighter areas indicate higher fat signal fraction. A 2 cm-sized circular ROI is placed in the right hepatic lobe, and the fat signal fraction for the ROI was calculated as 24.4%. MRS = MR spectroscopy, ROI = region of interest, TE = echo time

Several previous studies have shown that MRS is highly accurate in diagnosing hepatic steatosis using histologic results as a reference standard (area under the receiver operating characteristic curve [AUROC], 0.97–0.99 for detection of hepatic steatosis ≥ S1) [34,35]. However, MRS has an intrinsic limitation; it allows fat quantification of a small portion, usually a single voxel, of the liver, which may lead to sampling variability. Furthermore, MRS requires technical expertise for acquisition and analysis because of its complexity, which limits its widespread use [33,39].

Multi-Echo Dixon MRI

Fat quantification using multi-echo Dixon MRI is also based on the chemical shift phenomenon between fat and water protons. In gradient echo (GRE) imaging, the signals from fat and water periodically dephase and rephase owing to the difference between their precession frequencies [36,40]. When obtaining in-phase and out-of-phase signals, which is called a two-point Dixon method, a fat signal fraction can simply be estimated using the following equation:

S_IP = W + F, S_OP = W - F (if F < W)

$$\therefore \mathrm{F}\mathrm{a}\mathrm{t}\mathrm{\hspace{1em}}\mathrm{s}\mathrm{i}\mathrm{g}\mathrm{n}\mathrm{a}\mathrm{l}\mathrm{\hspace{1em}}\mathrm{f}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}=\frac{\mathrm{F}}{\mathrm{W}+\mathrm{F}}=\frac{\left|{\mathrm{S}}_{\mathrm{I}\mathrm{P}}-{\mathrm{S}}_{\mathrm{O}\mathrm{P}}\right|}{{2\mathrm{S}}_{\mathrm{I}\mathrm{P}}}$$

where S_IP and S_OP refer to signal intensities in the in-phase and out-of-phase images, respectively, and W and F refer to the signals from water and fat, respectively [41,42].

However, this method is limited by the dynamic range of the fat fraction from 0%–50%, and it can be affected by confounding factors, including T1 bias, T2^* decay, and the spectral complexity of fat [38]. Recent multi-echo Dixon techniques have overcome these confounding factors using a low flip angle, T2^* correction with multiple echoes, and multi-peak spectral modeling [38,43,44]. Using these multi-echo Dixon techniques, the signals from water and fat can be accurately separated and the fat signal fraction map of the entire liver can be obtained by calculating the signal ratio of the proton density of fat to the sum of those of fat and water (MRI-PDFF) (Fig. 2C) [38].

MRI-PDFF is highly accurate for the diagnosis and severity assessment of hepatic steatosis (AUROC, 0.98, 0.91, and 0.90 for ≥ S1, S2, and S3 in a meta-analysis, respectively) [45]. The diagnostic performance of MRI-PDFF is comparable to that of MRS (AUROC, 0.88 vs. 0.86 for ≥ S2) [46]; however, MRI-PDFF is more easily applicable because it does not require a technical expert for acquisition and analysis. MRI-PDFF has been widely used as a reference standard for performance studies on the quantification of hepatic steatosis and is the preferred endpoint for NASH clinical trials [16,17,39].

QUS Techniques

Although conventional B-mode ultrasound is used for a wide range of medical indications, quantitative information from B-mode ultrasound images is limited because ultrasound images are highly dependent on machine settings. However, recent technical developments allow ultrasound scanners not only to deliver images but also to obtain raw radiofrequency (RF) data, which enables the development of QUS [47]. QUS measures various acoustic parameters, including the attenuation coefficient (AC) [48], backscatter coefficient (BSC), speckle statistics [49,50], speed of sound [51,52], and elastography metrics [53,54] from the tissue, most of which are obtained from the raw RF data rather than processed images [47]. It aims to estimate tissue properties from these acoustic parameters by using appropriate models and theories of how ultrasound interacts with the tissue [47]. Since QUS can provide quantitative data related to tissue properties, it has been studied and utilized in various medical fields [49] such as the assessment of osteoporosis [55], characterization of the myocardium [56], characterization of breast and thyroid lesions [57,58,59], detection of prostate cancer and metastatic lymph nodes [60,61], and assessment of tumor response to chemotherapy [62,63], among others. In addition, QUS is expected to be effective in detecting hepatic steatosis, because the acoustic properties of hepatic tissue change with hepatic fat accumulation. Accordingly, multiple QUS techniques based on various acoustic parameters have been developed to quantitatively evaluate hepatic steatosis [64]. In this article, we introduce several representative QUS techniques based on AC, BSC, and speckle statistics for the evaluation of hepatic steatosis, which are briefly summarized in Table 1 and Figure 3.

Table 1. Summary of Quantitative Ultrasound Techniques for Evaluating Hepatic Steatosis.

	Attenuation		Backscatter (Including Speckle Statistics)
Physical meaning	Energy loss as ultrasound wave passes through tissue		Scatter that occurs when ultrasound wave strikes the microstructure of tissue
Ultrasound image appearance	Hypoechoic appearance at distant field in ultrasound image		Echogenic appearance in ultrasound image
Relationship with hepatic steatosis	Increasing with fat content		Increasing with fat content
Approach	AC (dB/cm/MHz), CAP (dB/m)		BSC, speckle statistics
Available software	Attenuation measurement without liver visualization		BSC
		- CAP (Echosens)		- Nothing commercialized
	Attenuation imaging using B-mode ultrasound		Speckle statistics
		- ATI (Canon Medical Systems)		- ASQ (Canon Medical Systems)
		- UGAP (GE Healthcare)		- TSI (Samsung Medison)
		- ATT (Hitachi)
		- Att PLUS (SuperSonic Imagine)
		- TAI (Samsung Medison)

AC = attenuation coefficient, BSC = backscatter coefficient, CAP = controlled attenuation parameter

Fig. 3. Schematic figures illustrating the physical meanings of quantitative ultrasound parameters.

Ultrasound wave, normal hepatocytes, and hepatocytes with fat accumulation are illustrated as arrows and hexagons with white and brown borders, respectively.

A. Ultrasound waves lose energy when they pass through the liver. Ultrasound waves lose more energy as they pass through fatty liver tissue than normal liver, resulting in a higher attenuation coefficient. B. Scattering occurs when an ultrasound wave hits the tissue microstructure of the liver. More scattering occurs in fatty liver tissue than in normal liver tissue, resulting in a higher backscatter coefficient. C. The speed of an ultrasound wave is slower in fatty liver tissue than in normal liver tissue.

Attenuation Coefficient (AC)

Attenuation refers to the energy loss when an ultrasound wave passes through tissue, and it is dependent on the tissue properties and the ultrasound frequency [64]. Ultrasound attenuation increases with hepatic fat infiltration, which obscures the hepatic vessels and diaphragm during conventional ultrasound [65,66,67]. AC is a quantitative measure of energy loss during ultrasound transmission [67]. There are two major approaches for the evaluation of hepatic steatosis using AC: 1) controlled attenuation parameter (CAP) obtained with the transient elastography device, using A-mode ultrasound and 2) B-mode ultrasound-guided attenuation imaging.

Controlled Attenuation Parameter (CAP)

CAP is one of the most widely studied QUS techniques for the quantification of hepatic steatosis, which uses an ultrasound-based vibration-controlled transient elastography (VCTE™) device (Fibroscan, Echosens). CAP is assessed simultaneously with liver stiffness measurement using raw RF data acquired by FibroScan [19]. To measure CAP, a patient should lie in the dorsal decubitus position with the right arm in maximum abduction. Then, an operator should place the appropriate probe on the intercostal space at the level of the right hepatic lobe [68]. Originally, a 3.5-MHz probe (M probe) was used to measure CAP, but a probe with a lower central frequency (XL probe, with a central frequency of 2.5 MHz) can be used with similar diagnostic performance, which can be useful for obese patients [69,70]. The probe should be placed in a portion of the liver with a > 6-cm thickness and without large vessels, and the placement can be assisted by ultrasound time-motion images. After the probe is placed at the appropriate site, acquisition of CAP and liver stiffness can be initiated by pressing the probe button [68]. The final CAP result is expressed as dB/m, which is correlated with the grade of hepatic steatosis [19]. The overall failure rate of CAP measurement using the M probe was reported to be 7.7%, which was associated with body mass index (BMI): 1.0% in patients with BMI ≤ 25 kg/m² and 58.4% in patients with BMI > 40 kg/m² [71]. The proper use of XL probes and automatic probe selection tools may reduce the failure rate [72].

The diagnostic performance of CAP has been variably reported as AUROCs ranging from 0.64 to > 0.90 (Table 2) [19,73,74,75,76,77,78,79,80,81,82,83]. In a meta-analysis of 19 studies involving 2735 patients, good overall diagnostic performance was reported as AUROCs of 0.823, 0.865, and 0.882 for the detection of hepatic steatosis grade ≥ S1, S2, and S3, respectively [84]. However, previous studies reported the inferiority of CAP to MRS (AUROC, 0.77 vs. 0.99 for ≥ S1) [34] or MRI-PDFF (AUROC, 0.88, 0.73, and 0.70 vs. 0.98, 0.90, and 0.79 for ≥ S1, S2, and S3, respectively) [85] for the diagnosis of hepatic steatosis.

Table 2. Summary of Studies Assessing Hepatic Steatosis Using CAP.

Study	N	Probe	Reference Standard	Target Degree of Steatosis	Optimal Cutoff (dB/m)	AUROC	Sen (%)	Spe (%)
Sasso et al. [19]	115	M	Biopsy	≥ S1 (10%)	238	0.91	91	81
				≥ S2	259	0.95	89	86
				≥ S3	292	0.89	100	78
de Lédinghen et al. [73]	112	M	Biopsy	≥ S1 (10%)	266	0.84	69	85
				≥ S2	311	0.86	57	94
				≥ S3	318	0.93	87	91
Myers et al. [74]	153	M	Biopsy	≥ 10%	283	0.81	76	79
				≥ S1 (5%)	289	0.79	68	88
				≥ S2	288	0.76	85	62
				≥ S3	283	0.70	94	47
Masaki et al. [75]	155	M	Biopsy	≥ S1 (5%)	233	0.88	87	77
Chan et al. [76]	161	M	Biopsy	≥ S1 (5%)	263	0.97	92	94
				≥ S2	263	0.86	97	68
				≥ S3	281	0.75	100	53
Chon et al. [77]	135	M	Biopsy	≥ S1 (5%)	250	0.89	73	95
				≥ S2	299	0.89	82	86
				≥ S3	327	0.80	78	84
Shen et al. [78]	152	M	Biopsy	≥ S1 (5%)	253	0.92	89	83
				≥ S2	285	0.92	93	83
				≥ S3	310	0.88	92	79
Karlas et al. [79]	65	M	Biopsy	≥ S1 (5%)	234	0.93	93	87
				≥ S2	269	0.94	97	81
				≥ S3	301	0.82	82	76
de Lédinghen et al. [80]	261	M	Biopsy	≥ S2	310	0.80	79	71
				≥ S3	311	0.66	87	47
Eddowes et al. [81]	380	M, XL	Biopsy	≥ S1 (5%)	302	0.87	80	83
				≥ S2	331	0.77	70	76
				≥ S3	337	0.70	72	63
Oeda et al. [82]	122	M	Biopsy	≥ S2	267	0.64	93	36
				≥ S3	286	0.69	90	43
		XL	Biopsy	≥ S2	273	0.68	96	32
				≥ S3	302	0.71	84	48
Caussy et al. [83]	100	M	MRI-PDFF	≥ 5%	294	0.84	75	78
				≥ 10%	311	0.89	79	85
		XL	MRI-PDFF	≥ 5%	307	0.86	74	75
				≥ 10%	322	0.93	83	87

AUROC = area under the receiver operating characteristic curve, CAP = controlled attenuation parameter, PDFF = proton density fat fraction, Sen = sensitivity, Spe = specificity

Nevertheless, CAP is less time-consuming and allows the simultaneous evaluation of steatosis and fibrosis [86,87]. It is also likely to be observer-independent with good interobserver agreement (concordance correlation coefficient, 0.82 between two raters) [88]. However, CAP can be affected by several other factors, including skin capsular distance [82,89] and probe type (M vs. XL probe) [16,83] and the cutoff value for the diagnosis of hepatic steatosis is poorly standardized and variable across studies (Table 2, Fig. 4A). In addition, CAP measurement from a sample volume is obtained blindly without a B-mode ultrasound image; therefore, the CAP value can be misevaluated due to the inadvertent inclusion of hepatic vessels, ducts, masses, or uneven steatosis [87].

Fig. 4. Scatter plots for a cutoff value of (A) controlled attenuation parameter and (B) attenuation coefficient measured with B-mode ultrasound-guided attenuation imaging for different degrees of hepatic steatosis.

A. Blue, red, and black dots indicate cutoff values when using M probe, XL probe, and both M and XL probes, respectively. B. Blue, red, and black dots indicate cutoff values when using ATI (Canon Medical Systems), UGAP (GE Healthcare), ATT (Hitachi), respectively. Only studies using liver biopsy as a reference standard are included.

B-Mode Ultrasound-Guided Attenuation Imaging

The measurement of AC under B-mode ultrasound guidance has been studied since the 1980s [65,66,90]. Recently, novel techniques for calculating the AC under B-mode ultrasound guidance have been commercialized for the evaluation of hepatic steatosis, including attenuation imaging (ATI; Canon Medical Systems) [20], ultrasound-guided attenuation parameter (UGAP; GE Healthcare) [21], attenuation coefficient (ATT; Hitachi) [22], and tissue attenuation imaging (TAI; Samsung Medison) [23]. Although the detailed evaluation method slightly differs between vendors, the general process of measurement is as follows: 1) B-mode ultrasound evaluation of the liver is performed using a convex probe, 2) the probe is located to visualize the right hepatic lobe through an intercostal window for AC measurement, 3) the region of interest (ROI) is placed in the right hepatic lobe at least 2 cm below the liver capsule to avoid reverberation artifacts during breath-hold while avoiding or automatically excluding large vessels, and 4) AC value (in dB/cm/MHz) and reliability of the measurement (in R²) are measured. A measurement of R² ≥ 0.60–0.90 is considered valid, depending on the vendors, and usually a median or mean value of five valid measurements is used for the assessment of hepatic steatosis (Fig. 5) [20,21,22,23,91]. The technical failure rate of these techniques, including ATI and UGAP, seems to be low (0%–4.3%), although there is little reported data [20,21,91,92,93].

Fig. 5. Various commercialized techniques for B-mode ultrasound-guided attenuation imaging.

A-D. ATI (Canon Medical Systems) of patients (A) without hepatic steatosis, and with (B) mild, (C) moderate, and (D) severe hepatic steatosis, which were confirmed by liver biopsy. Median ACs are measured as (A) 0.56, (B) 0.67, (C) 0.76, and (D) 0.86 dB/cm/MHz, respectively. E-H. UGAP (GE Healthcare) of patients (E) without hepatic steatosis, and with (F) mild, (G) moderate, and (H) severe hepatic steatosis, which are estimated by controlled attenuation parameter. Median ACs are measured by (E) 0.59, (F) 0.67, (G) 0.77, and (H) 0.85 dB/cm/MHz, respectively. I-L. TAI (Samsung Medison) of patients (I) without hepatic steatosis and with (J) mild, (K) moderate, and (L) severe hepatic steatosis, which were confirmed by MRI-proton density fat fraction. Median ACs are measured by (I) 0.62, (J) 0.73, (K) 0.80, and (L) 0.97 dB/cm/MHz, respectively. AC = attenuation coefficient

In several recent studies, AC calculated with these techniques generally showed a good diagnostic performance for hepatic steatosis, with liver biopsy or MRI-PDFF as reference standards (AUROC, 0.76–0.98 with different techniques, reference standards, and target degree of steatosis) [20,21,22,91,92,93,94,95,96,97,98]. In addition, AC has been shown to correlate well with the degree of steatosis evaluated by histology or MRI-PDFF (r = 0.47–0.78) [20,21,22,91,92,93,94,95,96,97]. The detailed results of the studies on ATI, UGAP, ATT, and TAI are summarized in Table 3 and Figure 4B.

Table 3. Summary of Studies Assessing Hepatic Steatosis Using B-Mode Ultrasound-Guided Attenuation Imaging.

Study	N	Technique	Reference Standard	r	Target Degree of Steatosis	Optimal Cutoff (dB/cm/MHz)	AUROC	Sen (%)	Spe (%)
Bae et al. [20]	108	ATI (Canon)	Biopsy	0.66	≥ S1 (5%)	0.64	0.84	75	77
					≥ 10%	0.66	0.88	80	83
					≥ S2	0.70	0.89	86	81
					≥ S3	0.75	0.93	100	82
Tada et al. [98]	148	ATI (Canon)	Biopsy	No data	≥ S1 (5%)	0.66	0.85	68	88
					≥ S2	0.67	0.91	92	84
					≥ S3	0.68	0.91	100	75
Jeon et al. [92]	87	ATI (Canon)	MRI-PDFF	0.66	≥ 5%	0.59	0.76	88	62
					≥ 10%	0.65	0.88	85	72
Dioguardi Burgio et al. [93]	101	ATI (Canon)	Biopsy	0.58	≥ S1 (5%)	0.69	0.81	76	86
					≥ S2	0.72	0.89	96	74
Jesper et al. [94]	27	ATI (Canon)	Biopsy	0.65	≥ S2	0.64	0.98	90	94
					≥ S3	0.68	0.98	100	90
Tada et al. [95]	119	ATI (Canon)	MRI-PDFF	0.70	≥ S1 (5.2%)	0.63	0.81	68	86
					≥ S2 (11.3%)	0.73	0.87	79	91
					≥ S3 (17.1%)	0.75	0.94	93	89
Ferraioli et al. [91]	72	ATI-Pen (Canon)	MRI-PDFF	0.78	≥ 5%	0.69	0.90	79	96
		ATI-Gen (Canon)	MRI-PDFF	0.83	≥ 5%	0.62	0.92	81	96
Fujiwara et al. [21]	163	UGAP (GE Healthcare)	Biopsy	0.78	≥ S1 (5%)	0.53	0.90	81	87
					≥ S2	0.60	0.95	86	82
					≥ S3	0.65	0.96	80	90
Tada et al. [96]	126	UGAP (GE Healthcare)	MRI-PDFF	0.75	≥ S1 (5.2%)	0.60	0.92	86	89
					≥ S2 (11.3%)	0.69	0.87	83	81
					≥ S3 (17.1%)	0.69	0.89	97	71
Tamaki et al. [22]	351	ATT (Hitachi)	Biopsy	0.47	≥ S1 (5%)	0.62	0.79	72	72
					≥ S2	0.67	0.87	82	82
					≥ S3	0.73	0.96	87	89
Jeon et al. [97]	120	TAI (Samsung)	MRI-PDFF	0.66	≥ 5%	0.88	0.86	78	79
					≥ 10%	0.98	0.84	64	93

AUROC = area under the receiver operating characteristic curve, PDFF = proton density fat fraction, r = correlation coefficient, Sen = sensitivity, Spe = specificity

The advantage of these techniques over CAP is their use of B-mode ultrasound images. First, conventional ultrasound evaluation of the liver can be performed simultaneously with fat quantification. Second, the ROI for calculating AC can be placed while visualizing the liver, and a more reliable result can be obtained by avoiding large vessels, ducts, and hepatic masses or cysts [20,21,22]. Studies have shown that ATI and UGAP are superior to CAP for the prediction of hepatic steatosis [21,91]. In addition, ATI, UGAP, and TAI showed high intra- and inter-observer reproducibility (intraclass correlation coefficients [ICCs] for intra-and inter-observer reproducibility, 0.93 and 0.79 for ATI, 0.86 and 0.84 for UGAP, and 0.99 and 0.99 for TAI, respectively) [21,23,99]. However, AC can also theoretically be affected by fibrosis, although the effect of fibrosis is less pronounced than steatosis [20]. Different results have been reported on the effects of hepatic fibrosis on AC measured with ATI, UGAP, or TAI [92,93,97,100,101]. Therefore, further studies and standardization of AC, with consideration of concurrent hepatic fibrosis, are warranted.

Backscatter Coefficient (BSC)

BSC is a quantitative measure of ultrasound energy reflected from a tissue during ultrasound examination and is related to the echogenicity or “brightness” of the tissue in conventional ultrasound. As echogenicity increases with fatty liver in conventional ultrasound, BSC is also known to increase with hepatic fat infiltration [66,67]. In some recent studies, BSC correlated well with the degree of hepatic steatosis evaluated by liver biopsy (r = 0.67) [67] or MRI-PDFF (r = 0.72 and 0.80) [67,102]. BSC has also been reported to have a good diagnostic performance for hepatic steatosis (AUROC, 0.85 and 0.83 for ≥ S2 and ≥ S3 and 0.95 for MRI-PDFF ≥ 5%) [67,102], with biopsy or MRI-PDFF as reference standards. However, these studies were in the research stage, which required post-processing of QUS data using a custom software.

Ultrasound Envelope Statistic Parametric Imaging (Speckle Statistics)

Ultrasound images contain speckle patterns that appear in a granular form. Since the speckle pattern is generated by the scattering of ultrasound signals by microstructures in the tissue, speckle statistics with the backscatter envelope can describe the scattering characteristics of the tissue [49,50,87]. The Rayleigh distribution generally describes the envelope of the backscattered ultrasound signal, which corresponds to the distribution of the envelope in the case of a high density of random scatterers without a coherent signal component [103,104]. However, because the distribution of the scattered ultrasound signals within the actual tissue does not always follow the Rayleigh distribution, various statistical models have been proposed [103,104,105,106,107]. Acoustic structure quantification (ASQ) and the Nakagami distribution have been the most widely studied for tissue characteristics.

Acoustic Structure Quantification (ASQ)

ASQ (Canon Medical Systems) is a quantification method for liver tissue characterization that measures the difference between the theoretical and real envelope distributions [108]. In ASQ, envelopes are used to compute Cm² by comparing the variance of the theoretical Rayleigh distribution and the real backscatter envelope distribution. Using limited envelope signals less than µ + 4σ, where µ and σ denote the mean and standard deviation of the envelope distribution, respectively, C_m² is recalculated as rC_m². The recalculated rC_m² and the original C_m² are compared to derive the focal disturbance ratio (FD ratio) [24,50,109]. In fatty liver, the echogenicity of the hepatic parenchyma is increased, and the hepatic vessel walls are blurred due to reflection and scattering of the ultrasound waves, which results in the homogenization of the signal strength [24]. Therefore, the FD ratio theoretically decreases in fatty liver [24].

The process of performing ASQ examination is as follows. First, B-mode ultrasound evaluation of the liver is performed. Next, ultrasound images in ASQ mode are acquired from the right intercostal and right subcostal view 3–5 times each. Display depth and transmit focus are set to 10 cm and 6 cm, respectively. Then, ROIs that are as large as possible are placed on the liver in the images, while avoiding large hepatic vessels and artifacts. Finally, the FD ratio is calculated automatically within the ROI and displayed on the monitor. The mean FD ratio can be used for analysis of hepatic steatosis [108,110]. The FD ratio measured in the intercostal and subcostal views did not show a significant difference and showed good agreement (ICC, 0.90) [108].

In early animal and human studies, the FD ratio measured by ASQ correlated well with the fat droplet area on biopsy (r = -0.75 to -0.72) [24,111] or MRS (r = -0.90 to -0.87) [108,110,112]. One study also showed good diagnostic performance of the FD ratio (AUROC, 0.96 for hepatic steatosis ≥ 10%) [108]. However, another clinical study showed a relatively weak correlation between the FD ratio and MRS (r = -0.43) and fair diagnostic performance of the FD ratio for the diagnosis of hepatic steatosis, defined by a CAP value of > 300 dB/m (AUROC, 0.76) [113]. Furthermore, there have also been several studies on the relationship between FD ratio and fibrosis, although the results are controversial, which can be a confounding factor when evaluating hepatic steatosis using ASQ [112,113,114,115,116]. Further studies on both steatosis and fibrosis are needed.

Nakagami Imaging

The Nakagami distribution is a generalized statistical model for evaluating the scattering characteristics within a tissue [50,104]. The Nakagami parameter (m) of the distribution is a shape parameter that depends only on the shape of the envelope distribution. The Nakagami parameter encompasses most scattering conditions. For m < 1, the envelope statistics represent a small number of randomly distributed scatterers. When m = 1, the envelope statistic is a Rayleigh distribution and represents a large number of randomly distributed scatterers. When m > 1, the envelope statistics represent a large number of randomly distributed scatterers with additional periodic scatterers [50,104]. Therefore, the backscattering characteristics of liver steatosis can be explained by the Nakagami parameter with specific physical meanings according to the various amounts and spatial arrangement of scatterers.

Early animal and human studies revealed a significant positive correlation between the Nakagami parameter and the lipid concentration of the liver tissue (r = 0.86 and 0.79 for cholesterol and triglyceride, respectively) [117] and the degree of hepatic steatosis assessed by a conventional ultrasound-based scoring system (r = 0.84) [118].

Recently, a commercially available QUS modality based on the Nakagami distribution, tissue scatter-distribution imaging (TSI, Samsung Medison), was introduced (Fig. 6) [23,97,101]. The image acquisition process of TSI is similar to that of TAI. First, B-mode ultrasound images are acquired at the right hepatic lobe through the intercostal window near the level of the hepatic hilum. Then, a function key for the TSI is selected and an ROI box is generated. The operator should place the ROI in a relatively homogeneous region in the right hepatic lobe, at least 2 cm below the liver capsule. Large hepatic vessels, focal fat sparing or deposition, and artifacts should be avoided as for other QUS techniques, including TAI. Finally, the TSI parameter (TSI-p, which is equal to m × 100) is calculated and the mean or median values of TSI-p are used for the analysis of hepatic steatosis [23].

Fig. 6. A commercialized quantitative ultrasound technique based on Nakagami distribution–TSI (Samsung Medison).

A. TSI image (left) and corresponding histogram of the scattered ultrasound signals (right) in a patient without hepatic steatosis. Median TSI-p was calculated as 84.26. The histogram shows pre-Rayleigh distribution compared to the ideal Rayleigh distribution (red curve). B. TSI image (left) and corresponding histogram (right) in a patient with mild hepatic steatosis. Median TSI-p was calculated as 94.69. The histogram shows pre-Rayleigh distribution compared to the ideal Rayleigh distribution (red curve). C. TSI image (left) and corresponding histogram (right) in a patient with moderate hepatic steatosis. Median TSI-p was calculated as 100.44. The histogram is similar to the ideal Rayleigh distribution (red curve). D. TSI image (left) and corresponding histogram (right) in a patient with severe hepatic steatosis. Median TSI-p was calculated as 107.14. The histogram shows the post-Rayleigh distribution compared to the ideal Rayleigh distribution (red curve). Fat signal fractions measured with MRI-proton density fat fraction in these patients were (A) 3.1%, (B) 7.1%, (C) 12.6%, and (D) 25.2%, respectively.

TSI = tissue scatter-distribution imaging, TSI-p = TSI-parameter

In recent studies, the TSI-p showed a good correlation with both CAP (r = 0.68, with CAP value [23], and r = 0.59 with steatosis grade determined by CAP [101]) and MRI-PDFF (r = 0.73) [97]. TSI also showed excellent performance for the diagnosis of hepatic steatosis (AUROC, 0.96 for hepatic fat content ≥ 5% and 0.94 for hepatic fat content ≥ 10%), with MRI-PDFF as a reference standard [97] and good intra- and inter-observer agreements (ICC, 0.98 and 0.95, respectively) [23]. However, there are controversial results on the effect of TSI-p on fibrosis, which is another important pathological feature of NAFLD/NASH [97,101]. Therefore, further validation with consideration of fibrosis is warranted.

Discussion and Future Development

For the diagnosis of hepatic steatosis, MRI-PDFF (sensitivity, 93%; specificity, 94% for ≥ S1) [45] and MRS (sensitivity, 88.5%; specificity, 92.0% for ≥ 0%–5%) [27] are the most accurate among imaging modalities, while CT is much less sensitive than MR-based methods (sensitivity, 46.1%; specificity, 93.5% for ≥ 0%–5%) [27]. Although variably reported in different studies, CAP and B-mode ultrasound-guided attenuation imaging seems to be more sensitive than CT, but less accurate than MR-based methods (Tables 2, 3). Meanwhile, ASQ and TSI showed excellent performance with MRS or MRI-PDFF as reference standards in each single prospective study (ASQ, sensitivity, 86.2% and specificity, 100% for ≥ 10% [108]; TSI, sensitivity, 85.7% and specificity, 97.4% for ≥ 5% [97]). However, larger multicenter studies are needed to validate these findings.

Although MRI-PDFF and MRS are the most accurate, reproducible, and well-validated methods for liver fat quantification, they are not routinely used for NAFLD screening because of their limited accessibility and low costeffectiveness [38,119]. Considering the growing incidence of NAFLD, a more available, cost-effective, and easy-to-operate noninvasive diagnostic tool for the evaluation of hepatic steatosis is needed. Ultrasound is recommended as the first-line diagnostic method for assessing steatosis; however, it is limited by its relatively low sensitivity and substantial intra- and inter-observer variability [27,28]. Meanwhile, QUS generally showed good intra- and inter-observer agreements regardless of the technique used [21,23,88,99]. Furthermore, QUS provides continuous values related to hepatic fat content, unlike conventional ultrasound, which can provide only subjective categorical values; this can be useful for longitudinal follow-up and evaluation of treatment response [97]. In this context, QUS is a promising tool for screening and treatment monitoring of patients with NAFLD. In addition to NAFLD, QUS can potentially be applied to any condition where hepatic fat accumulation affects the prognosis of patients. For example, steatosis of ≥ 30% in a liver graft is associated with an increased risk of graft loss after liver transplantation [120]. In addition, the severity of hepatic steatosis is associated with patient outcomes and mortality after liver surgery [121]. Therefore, QUS techniques can potentially be used as a non-invasive preoperative or pre-transplantation evaluation tool for the presence and degree of hepatic steatosis. Various QUS techniques, including CAP [19], attenuation imaging [20,21,22,23], ASQ [24], and Nakagami imaging [23], have been commercialized, and have shown promising results for quantitative evaluation of hepatic steatosis, although further validation and standardization between vendors or platforms are needed for clinical adoption.

Inflammation and fibrosis are also important histologic features of NAFLD, which can affect the treatment strategy [122]. Although transient elastography is a well-validated method for the evaluation of hepatic fibrosis [123,124,125], it is limited by blinded evaluation without B-mode ultrasound guidance. ASQ has been studied for the evaluation of fibrosis; however, its performance, as mentioned earlier, is controversial [113,114,115,116]. Thus, there is a need for noninvasive evaluation of inflammation or fibrosis in patients with NASH/NAFLD. Recently, shear-wave elastography and shear-wave dispersion imaging (viscosity imaging) based on ultrasound have shown good outcomes for detection of fibrosis [126,127] and inflammation [128,129], respectively. These techniques, in conjunction with QUS techniques for hepatic fat quantification, may enable comprehensive evaluation of patients with NASH/NAFLD using ultrasound. Further studies validating these imaging-based biomarkers in a large independent group are needed.

Availability of Data and Material

Data sharing does not apply to this article as no datasets were generated or analyzed during the current study.