Methods and Validation of Velacur Determined Fat Fraction in patients with MASLD

Abstract

Introduction:

As prevalence of patients with steatotic liver diseases increases throughout the world, it is necessary to have accurate and accessible methods to estimate liver fat content. Using quantitative ultrasound parameters, such as attenuation and backscatter, it is possible to estimate liver fat, with MRI proton density fat fraction as the reference standard. Velacur determined fat fraction (VDFF) is a new output measurement on Velacur (Sonic Incytes Medical Corp, Vancouver, BC).

Methods:

This study described the results of parameter fitting and validation of VDFF, which is a combination of quantitative ultrasound parameters. Patients were recruited from sites within the US and Canada. All patients had contemporaneous Velacur and MRI proton density fat fraction scans. The quantitative ultrasound parameter fitting was completed using linear regression on a random sub-sample approach, and a separate cohort was used for validation. The AUC for detection of 5% liver fat based on MRI-PDFF and the correlation between MRI-PDFF and VDFF was measured in both cohorts.

Results:

VDFF had an AUROC of 0.97 for the detection of MRI-PDFF > 5% in the parameter fitting cohort, and 0.99 in the validation cohort. The correlation [95% CI] between MRI-PDFF and VDFF was r = 0.84 [0.78 – 0.89] for the parameter fitting cohort and r = 0.90 [0.82 – 0.95] for the validation cohort.

Conclusion:

The Velacur Determined Fat Fraction (VDFF) is an accurate and accessible way to estimate steatosis as measured by MRI-PDFF. Velacur VDFF can fill the unmet need of an accurate means to diagnosis hepatic steatosis and serve as a potential alternative to biopsy or MRI-PDFF.

1. Introduction

Liver biopsy remains the current gold standard for the diagnosis of many chronic liver diseases including Steatotic Liver Disease (SLD), Metabolic-dysfunction Associated Steatotic Liver Disease (MASLD) and Metabolic-dysfunction Associated Steatohepatitis (MASH). Although it is the current gold standard, liver biopsy has risks of complications [1], [2] and misclassification due to sampling error and inter-observer and intra-observer variation [3], [4]. With the growing prevalence of SLD in the general population, and especially in patients with diabetes, there is a need for more accessible and accurate non-invasive methods for steatosis evaluation [5], [6], [7]. New pharmaceutical options that will soon be available for SLD, MASLD and/or MASH are also accentuating the unmet need for assessment of hepatic steatosis [8].

MRI Proton Density Fat Fraction (MRI PDFF) has emerged as the leading non-invasive method for steatosis assessment [9], [10], [11], [12], [13]. This method is able to measure the amount of water vs triglyceride protons within the tissue to estimate the overall fat fraction. Although highly accurate, this method is not accessible at the point of care for most patients and, given the high burden of disease in the general population, is not practical for patient screening.

Multiple society guidelines advocate for the use of non-invasive tests (NITs) for the assessment of patients with SLD or for those who are at risk of MASLD and MASH based on the presence of diabetes and/or metabolic syndrome [1], [14], [15], [16]. The 2023 American association for the study of liver diseases (AASLD) Practice Guidance on the clinical assessment and management of non-alcoholic fatty liver disease specifically outlines a tiered approach of assessment using both blood-based markers and imaging of liver stiffness and fat [1]. The liver society guidance documents commonly cite > 5% liver fat on MRI-PDFF as the definition of steatosis [1], [14], [15], [16]. This means that the accurate detection of greater than 5% liver fat is an essential part of the diagnosis of liver steatosis.

Ultrasound is an accessible modality with the promise to quantify hepatic steatosis [17], [18], [19], [20], [21], [22]. Quantitative ultrasound (QUS), or measuring parameters that can be derived directly from the ultrasound signal, separate from the created image, has been used for this purpose [22]. Ultrasound attenuation has been shown to correlate with liver fat from biopsy [21], [23], [24] and with MRI-PDFF [25], [26], [27]. Parameters of qualitative ultrasound, such as backscatter coefficient and attenuation, can be further combined to better correlate with MRI-PDFF. This has been shown in the literature [22], [28] and a method of combining quantitative ultrasound parameters to estimate liver fat is commercially available on the Siemens Acuson Sequoia system (Ultrasonically Derived Fat Fraction, UDFF) [28].

Quantitative ultrasound methods have also been compared with MRI-PDFF for both adults and children, including Siemens UDFF [20], [28], FibroScan^® Controlled Attenuation Parameter (CAP) (EchoSens) [26], backscatter alone [17] and other quantitative ultrasound parameter combinations [29].

Velacur^® is a commercially available point of care ultrasound-based device that is specifically designed to be used for liver assessment [30] (Figure 1). The commercially available system was designed to incorporate both liver stiffness and attenuation measurements. An abdominal ultrasound probe is swept over the right lobe of the liver to collect a volume of measurements up to 15cm in depth. The aim of this study was to define a third output for Velacur, the estimate of hepatic steatosis known as Velacur Determined Fat Fraction (VDFF). VDFF uses quantitative ultrasound measurements to estimate hepatic steatosis.

Figure 1:

Velacur^® is a commercially available point of care ultrasound-based device consisting of a user interface, an activation unit (to produce shear waves), and ultrasound probe and a control unit (A). The abdominal ultrasound probe is manually swept over the liver to collect measurements over a large volume of the right lobe (B). Measurements are collected up to 15cm in depth.

2. Methods

2.1. Study Design:

Two cohorts of patients were included in this analysis. The first cohort included patients consecutively or conveniently recruited at 4 sites in the US and Canada (University of California San Diego, Southern California Research Center, Beth Israel Deaconess Medical Center, University of British Columbia). This cohort is used to determine the VDFF parameter coefficients and will be referred to as the parameter fitting cohort. A second cohort of patients recruited from two separate sites in the US (Washington University School of Medicine) and Canada (Vancouver Coastal Health Research Institute) was used as validation, and will be referred to as the validation cohort.

All participants had suspected or diagnosed MASLD or MASH at the time of enrollment and underwent both the Velacur and MRI scans. Data obtained during Velacur scans performed on participants in the Parameter fitting cohort was used to calibrate and determine the coefficients of the attenuation and backscatter parameters used to calculate VDFF. The study was conducted in accordance with Good Clinical Practices. Informed consent, in writing, was obtained from each patient before their participation in the study and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the appropriate institutional review committees.

2.2. Inclusion and Exclusion Criteria:

The parameter fitting cohort consisted of participants between the ages of 19 and 75 years. Only patients with MASLD and/or suspected MASH were enrolled.. Excluded patients included those with viral hepatitis, other known causes of chronic liver disease, decompensated cirrhosis, serum ALT or AST > 5 × ULN on historical blood work within the past 3 months, individuals with history of persistent ethanol abuse, or patients who were pregnant or planning to become pregnant during the study. Persistent ethanol use was defined as individuals with history of persistent ethanol abuse (consumption > 20 g ethanol/day for women, > 40 g ethanol/day for men) for more than 3 months in the past year. Patients with BMI greater than 40 kg/m² (or as determined by the MRI bore size) were also excluded.

The validation cohort consisted of patients with the above inclusion criteria (6/38) and participants recruited at the Washington University School of Medicine, St. Louis, MO (32/38). Those from St. Louis were aged between 18–70 years of age, with a BMI no greater than 50 kg/m² and included people with and without MRI-diagnosed hepatic steatosis.

Participants in this study were allowed to choose as many race/ethnicity options as they felt described themselves. The percentage white (Table 1), is the percentage of participants who chose only white as a description of their race/ethnicity. The mixed race percentage was the percentage of participants who chose white and another category. Blood work results were collected from the patient’s chart of their most recent visit.

Table 1.

Summary of Patient Demographic Information

Characteristic	Parameter Fitting Cohort	Validation Cohort
Included Subjects	113	38
Sex (% Female)	51.3%	84%
Age (mean ± std)	57.4 ± 11.7 years	46.5 ± 14
BMI (mean ± std)	30 ± 4.3 kg/m2	30 ± 7.9
Race/Ethnicity (% white)	37.1%	73.7%
% mixed race	6.2%	0%
T2DM (% with)	15.0%	15.8%
Platelets (×109)	216.0 ± 72.9	254.1 ± 57.6
AST (U/L)	41.3 ± 22.2	23.6 ± 11.5
ALT (U/L)	52.2 ± 30.4	26.7 ± 22.8
MRI-PDFF (% mean ± std)	14.0 ± 8.2	8.3 ± 8.2
with MRI-PDFF <5%	3.4%	2.4%
with MRI-PDFF 5–11%	7.9%	7.4%
with MRI-PDFF 11–17%	14.1%	12.5%
with MRI-PDFF >17%	23.4%	23.2%
VDFF (% mean ± std)	13.4 ± 6.6	10.7 ± 6.6
with MRI-PDFF <5%	4.9%	4.8%
with MRI-PDFF 5–11%	11.0%	10.8%
with MRI-PDFF 11–17%	14.9%	13.6%
with MRI-PDFF >17%	20.2%	21.0%

Abbreviations: ALT, alanine transaminase; AST, aspartate aminotransferase; BMI, body mass index; T2DM, type 2 diabetes mellitus.

2.3. MRI-PDFF:

MRI-PDFF was used as the reference standard for liver fat fraction for this study. MRI-PDFF is available on most MRI systems. The standard MRI-PDFF protocol for each manufacturer was used. All scans were taken contemporaneously, with MRI’s taken within 28 days of the Velacur scans.

MRI scans were segmented manually by a radiologist or by AMRA^®. The radiologist segmented multiple areas of at least 2cm in diameter through out the volume of the right lobe of the liver. Segmenters were instructed to avoid vessels and areas near the liver boundary. The mean of all segmented areas or the reported AMRA^® result was used as the final result for each subject.

2.4. Velacur Determined Fat Fraction (VDFF):

VDFF is a measurement derived from quantitative measurements calculated from the radio frequency data collected during the Velacur scans. Velacur is a commercially available system to measure ultrasound elastography and ultrasound attenuation. As part of the measurement of tissue displacements for stiffness measurements, Velacur utilizes a method of high-speed imaging to measure each block of tissue multiple times. The multiple measurements (up to 25 measurements per line) allow for many more repeated measures to be used in the quantitative assessment of tissue. These repeated measurements allow for more data to be averaged during the calculation of each parameter in each location creating better overall estimates. Because of this, Velacur is uniquely suited to accurately measure quantitative ultrasound features such as attenuation and backscatter. The averaged data for each ultrasound parameter, for each patient is used as the input for the VDFF parameter fitting. In addition to the ability to average multiple lines, Velacur incorporates the use of an algorithm to define the location of liver and shear waves within the images [31], [32]. Only data from the liver and with high quality shear waves, as defined by the Velacur system quality algorithms, was used in the analysis.

VDFF is a linear combination of backscatter (BSC) and attenuation (ACE) parameters calculated from the collected Velacur ultrasound data [33]. Attenuation measures the energy loss as the ultrasound wave travels through tissue and is higher in fatty tissue. Backscatter is a measure of the amount of scattering which occurs as the ultrasound waves interacts with tissue microstructure. Tissue with high levels of backscatter will look brighter in the B-Mode image and this increase in brightness is a marker of liver steatosis as seen by sonographers [33]. Each of these parameters was independently calibrated and validated against phantoms of known values before being applied to this patient data set.

The contribution of each parameter to the final VDFF result was defined using linear regression on a repeated random sub-sample of the parameter fitting cohort. In this case, multiple rounds of fitting were used. During each round, 80% of the data was included and linear regression was used to find the BSE and ACE coefficients, using PDFF as the reference standard. This was repeated 1000 times. The distribution of the coefficients from each round was examined to see if there was any large deviations caused by outliners in the patient cohort. As the coefficients were normally distributed (verified using one-sample Kolmogorov-Smirnov test), the mean value of each coefficients was used as the final coefficient in the linear combination of BSC and ACE. The final calculated VDFF was then applied to all the participants in both the parameter fitting and validation cohorts and reported below.

2.5. Study Objectives:

The primary objective of this study was to determine the accuracy of VDFF in patients with MASLD, using the measure 5% MRI-PDFF as the reference standard. The secondary objective was to determine the accuracy at higher levels of MRI-PDFF, at 11% and 17%. These cut-off points were chosen to approximate the histological steatosis grades. Although the MRI-PDFF cut-off points that correspond to histology steatosis grades differ across studies, these cut-offs were defined by Imajo et al. [23]. In addition, the overall correlation coefficients between MRI and VDFF are reported.

The VDFF parameters are first set using the MRI-PDFF measurements from patients within the parameter fitting cohort. The final VDFF output is then tested against the MRI-PDFF measurements from the validation cohort.

An exploratory objective of the study was to compare the performance of CAP to VDFF in the parameter fitting cohort of patients, who also received contemporaneous FibroScan exams (within 28 days).

2.6. Statistical plan

Receiver operator curve (ROC) analysis is used to measure the performance of VDFF in the detection of MASLD (MRI-PDFF >5%) and more advanced disease (MRI-PDFF >11% and >17%). For each MRI-PDFF cut-off, the mean area under the curve (AUC) and 95% confidence intervals (CI), sensitivity, specificity are calculated. This cut-offs are based on those reported by Imajo et al. [23]. The F1 metric is reported for each cut-off.

Pearsons correlation coefficient are used to measure the overall correlation between VDFF and MRI-PDFF.

Results from both the parameter fitting and validation cohorts are reported.

3. Results

3.1. Patient Characteristics:

A total of 113 participants with paired MRI-PDFF and Velacur scans were included in the parameter fitting cohort and 38 participants with paired scans were included in the validation cohort. The overall participant characteristics are outlined in Table 1. Participants in the parameter fitting cohort were 51% female and 37.1% white, with a mean age of 57.4 year and mean body mass index (BMI) of 30 kg/m². Fifteen percent (17/113) had an MRI-PDFF of less than 5%, 26% (30/113) had a MRI-PDFF of 5–11%, 24% (28/113) from 11–17% and 34% (38/113) with a MRI-PDFF of 17% or greater.

The validation cohort was more homogeneous, with 73.7% white and 84% female. In addition, there were a higher percentage of participants (50%) with MRI-PDFF less than 5%. The average MRI-PDFF and VDFF were significantly lower in the validation cohort than the parameter fitting cohort, at 8.3% vs 14% for PDFF, and 10.7% and 13.4% for VDFF respectively (p < 0.05 using unpaired t-test).

3.2. Detection of MASLD:

3.2.1. Parameter Fitting Cohort

The distribution of patients into each MRI-PDFF grouping is shown in Table 2. VDFF had an AUROC of 0.97 for the detection of MRI-PDFF > 5%. The resulting AUC (mean [95% CI]) for the other MRI cut-off of 11% and 17% was 0.91 [0.85 – 0.95] and 0.91 [0.83 – 0.95] respectively. Overall, the AUC for VDFF is excellent at each of the MRI cut-offs. The sensitivity and specificity of 5% VDFF was 98% for VDFF and the specificity was 41%.

Table 2.

Parameter Fitting Cohort. Resulting AUC for each of the MRI cutoffs. 5% defines the diagnosis of any steatosis. Using the VDFF values (5%, 11% and 17%), the sensitivity and specificity of the VDFF cut points is shown for the parameter fitting cohort.

MRI-PDFF Cutoff	Number of Patients	AUROC [95% CI]	Sensitivity	Specificity	F1 Metric
< 5%	17
5%	30	0.97 [0.92 – 0.99]	98%	41%	0.94
11%	28	0.91 [0.85 – 0.95]	92%	70%	0.87
17%	38	0.91 [0.83 – 0.95]	82%	87%	0.78

3.2.2. Validation cohort

Table 3 outlines the results of VDFF when applied to this validation cohort. The overall AUC’s are similar to or greater than those of the parameter fitting cohort.

Table 3.

Validation Cohort. Resulting AUC for each of the MRI cut-offs. 5% defines the diagnosis of any steatosis. Using the VDFF values (5%, 11% and 17%), the sensitivity and specificity of the VDFF cut points is shown for the validation cohort.

MIR-PDFF Cutoff	Number of Patients	AUROC [95% CI]	Sensitivity	Specificity	F1 Metric
< 5%	19
5%	8	0.997 [0.97 – 1.00]	100%	47%	0.81
11%	4	0.97 [0.86 – 1.00]	91%	85%	0.80
17%	7	0.98 [0.86 – 1.00]	100%	97%	0.93

3.3. Correlation Between VDFF and MRI-PDFF

The overall correlation between MRI-PDFF and VDFF was 0.84 [0.78 – 0.89] for the training cohort (Figure 2 left) and 0.90 [0.82 – 0.95] for the validation cohort (Figure 2 right).

Figure 2:

Scatter plots of MRI-PDFF and VDFF outputs. The results of the parameter fitting cohort (left) and the validation cohort (right).

3.4. Comparison to FibroScan CAP:

As an exploratory objective of the study, the results for FibroScan CAP were also analysed in the parameter fitting cohort (the validation cohort did not have FibroScan data available). In line with other literature discussed below, the CAP AUC was 0.85 [0.76, 0.92], 0.80 [0.71, 0.87] and 0.81 [0.71, 0.89] for the MRI cut-offs of 5%, 11%, and 17% respectively. Using DeLong’s test to test for significance, shows that all the AUC’s are significantly different (ρ < 0.05) and higher for VDFF than CAP [34].

The overall correlation between CAP and MRI-PDFF in this cohort was 0.60 (0.47–0.71), which is also lower than the correlation between VDFF and MRI-PDFF (0.84 [0.78–0.89]).

4. Discussion

In this cohort of patients with a broad range of MRI-PDFF values, the Velacur Determined Fat Fraction (VDFF) was shown to be an accurate estimation of MRI-PDFF, with AUC’s greater than 0.9 in both the parameter fitting and validation cohorts, at all levels of MRI-PDFF. VDFF has the potential to be used in point of care settings, providing an alternative to MRI-PDFF in the diagnosis of patients with MASLD and MASH. As the number of patients in the US and around the world increases, it is important to have options for screening and diagnose patients with hepatic steatosis in order to provide correct and timely interventions. Neither MRI-PDFF nor biopsy are able to scale to the degree needed.

Although both attenuation and backscatter, as well as speed of sound, have been shown to correlate with MRI-PDFF and hepatic fat as measured from biopsy, VDFF adds to the current literature. Velacur offers a novel method of both collecting multiple measurements, and the collection of many more samples at the same location through the use of the high-speed imaging used.

This results presented here are as good as or better than those previously reported in the literature. Paige et al. reviewed a number of studies using ultrasound attenuation from several manufacturers, and CAP specifically, using MRI-PDFF or biopsy as the reference [18]. The average AUC in this study for all steatosis cut-offs across all listed studies using ultrasound attenuation was about 0.89, and 0.83 for CAP. The UDFF results showed an AUC for detection of 5% steatosis of 0.75, vs conventional ultrasound at 0.53 [28]. In pediatrics, the UDFF AUC for 6% fat on MRI was 0.95 [20]. In a direct comparison between one specific QUS combination and CAP using MRI-PDFF as the reference standard, QUS had a significantly higher AUC for diagnosis of MASLD (5% on MRI PDFF) than CAP (0.92 vs 0.79) [29]. Using these past results as basis for comparison, the VDFF compares favourability, with an average AUC of 0.93 in this cohort of patients.

It is interesting to note that the results in the validation cohort where actually better than those from the parameter fitting cohort. This may be due to the fact that the validation was very homogeneous, or that there were more patients with lower MRI-PDFF values.

As more and more quantitative ultrasound techniques are developed, it is important to be able to compare and standardize the measurements for multiple techniques. There are several initiatives, including those spearheaded by the Quantitative Imaging Biomarker Alliance (QIBA), to standardize the measurements of attenuation, backscatter and speed of sound across multiple ultrasound manufacturers [35], [36], [37].

4.1. Limitations

Although the parameters for VDFF were fitting using a random sub-sample approach, and validated on a small external cohort, further validation of the measure is needed using a more diverse cohort of patients and on patients from other backgrounds.

Fatty Liver, MASLD and MASH, are a global issue, and an issue within North America, with estimates ranging from 20–30%. Patients in this study were recruited from several sites, but all these sites were within North America and may include bias due to their origin. In particular, the validation cohort was very homogeneous in sex and race. This may limit the generalizability of the method, and further validation on cohorts of patients with more diverse backgrounds are needed. The parameter fitting cohort, although more diverse were still all recruited within North America. We look forward to understanding the performance of the algorithm in different populations.

As both cohorts were relatively small, there were concerns about overfitting and underfitting. The validation cohort was very different, in sex, age, race and distribution of steatosis. As the performance of VDFF was still good within this, very different, population, we believe that over and underfitting are not a concern but will complete further validation of VDFF.

5. Conclusions

The Velacur Determined Fat Fraction (VDFF) is an accurate and accessible way to estimate MRI-PDFF in patients with a wide range of MRI-PDFF. The high AUC’s for detection of MRI-PDFF cut offs (>0.9) show that the VDFF is an accurate estimate of liver fat. Velacur, and other quantitative ultrasound methods, can fill the unmet need of an accurate means to diagnosis hepatic steatosis. As an accessible, non-invasive measurement, VDFF could be used as an alternative to biopsy or MRI-PDFF

Further studies should further validate VDFF in additional patient cohorts and other disease populations. When measured in patients undergoing lifestyle modification or therapeutic interventions VDFF should show changes over time that correlation with MRI-PDFF responses.

Abbreviations:

AUROC

Area under receiver operator curve

ACE

Attenuation Coefficient Estimate

ALT

Alanine transaminase

AST

Aspartate aminotransferase

BMI

Body mass index

CAP

Controlled Attenuation Parameter

CI

Confidence Interval

MASH

Metabolic associated steatohepatitis

MASLD

Metabolic dysfunction associated liver disease

MRE

Magnetic resonance elastography

MRI

Magnetic resonance imaging

MRI-PDFF

Magnetic resonance imaging proton density fat fraction

NIT

Non-invasive test

QIBA

Quantitative Imaging Biomarker Alliance

QUS

Quantitative Ultrasound

SLD

Steatotic Liver Disease

VDFF

Velacur Determined fat Fraction

UDFF

Ultrasonically derived fat fraction