Published online May 23, 2019.
https://doi.org/10.3348/kjr.2018.0812
Technical Performance of Two-Dimensional Shear Wave Elastography for Measuring Liver Stiffness: A Systematic Review and Meta-Analysis
Abstract
Objective
To assess the technical performance of two-dimensional shear wave elastography (2D-SWE) for measuring liver stiffness.
Materials and Methods
The Ovid-MEDLINE and EMBASE databases were searched for studies reporting the technical performance of 2D-SWE, including concerns with technical failures, unreliable measurements, interobserver reliability, and/or intraobserver reliability, published until June 30, 2018. The pooled proportion of technical failure and unreliable measurements was calculated using meta-analytic pooling via the random-effects model and inverse variance method for calculating weights. Subgroup analyses were performed to explore potential causes of heterogeneity. The pooled intraclass correlation coefficients (ICCs) for interobserver and intraobserver reliability were calculated using the Hedges-Olkin method with Fisher's Z transformation of the correlation coefficient.
Results
The search yielded 34 articles. From 20 2D-SWE studies including 6196 patients, the pooled proportion of technical failure was 2.3% (95% confidence interval [CI], 1.3–3.9%). The pooled proportion of unreliable measurements from 20 studies including 6961 patients was 7.5% (95% CI, 4.7–11.7%). In the subgroup analyses, studies conducting more than three measurements showed fewer unreliable measurements than did those with three measurements or less, but no intergroup difference was found in technical failure. The pooled ICCs for interobserver reliability (from 10 studies including 517 patients) and intraobserver reliability (from 7 studies including 679 patients) were 0.87 (95% CI, 0.82–0.90) and 0.93 (95% CI, 0.89–0.95), respectively, suggesting good to excellent reliability.
Conclusion
2D-SWE shows good technical performance for assessing liver stiffness, with high technical success and reliability. Future studies should establish the quality criteria and optimal number of measurements.
INTRODUCTION
Ultrasound (US) elastography is a non-invasive tool used in chronic liver disease for staging liver fibrosis or predicting portal hypertension. Among several US elastography techniques, two-dimensional shear wave elastography (2D-SWE) is the latest method using an acoustic radiation force impulse (ARFI) to cause liver-tissue deformation and eventually generate a shear wave. It provides a 2D quantitative map of liver stiffness values over a large region of interest (ROI) by placing the ARFI focus at multiple sequential locations and capturing the generated shear waves. Because 2D-SWE involves real-time imaging, both the depth and size of sampling areas can be chosen manually at desired locations with no mass, large vessels, or artifacts. 2D-SWE has been integrated into most clinical US systems with the same probes as that used in traditional US (1).
Owing to its advantages, 2D-SWE helps assess the stability of measuring and quantifying an average stiffness value in a large ROI for higher reliability (2). However, because of its relative novelty, 2D-SWE has not yet been validated and some aspects remain incompletely clarified (3). Validating a diagnostic device for clinical use involves two main processes: 1) diagnostic accuracy—the evidentiary process of linking a biomarker with clinical endpoints and biologic processes and 2) technical performance—assessment of technical success/failure and measurement variability (4).
Thus far, most clinical validation attempts have focused on the good diagnostic accuracy of 2D-SWE for the degree of liver fibrosis (1, 2, 5, 6). Nevertheless, its technical performance also needs assessment. Although 2D-SWE systems from different manufacturers have custom built-in indicators for better measurement quality and stability, the evidence supporting them are limited (7). Indeed, previous studies evaluating the technical performance of 2D-SWE were generally small-scale studies with low-level evidence (8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41). To increase the level of evidence and arrive at more evidence-based results, sufficient evidence should be accumulated and summarized.
Therefore, we conducted this systematic review and meta-analysis to evaluate the technical performance of 2D-SWE for measuring liver stiffness.
MATERIALS AND METHODS
Institutional Review Board approval was not required because of the nature of our study, which was a systemic review and meta-analysis. Our systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (42).
Literature Search Strategy
We conducted an electronic literature search to identify suitable studies from the Ovid-MEDLINE (U.S. National Library of Medicine) and EMBASE (Elsevier) databases until June 30, 2018 (Supplementary Materials in the online-only Data Supplement).
Eligibility Criteria and Study Selection
We tried to evaluate the technical performance of 2D-SWE for measuring liver stiffness. Thus, we included studies and study subsets that evaluated any of the following outcomes by using 2D-SWE for measuring liver stiffness: 1) technical failure; 2) unreliable measurements; 3) interobserver reliability; and 4) intraobserver reliability.
Technical failure was the inability to obtain an adequate signal for all acquisitions, which was adopted in all studies consistently. As unreliable measurements were randomly defined across studies without consensus, we used the slightly different definitions of unreliable results employed in each of the included studies. Regarding measurement reliability, we included studies comparing the stiffness between different observers (interobserver reliability) and between different sessions by the same observer (intraobserver reliability).
The exclusion criteria were as follows: 1) studies reporting insufficient data for outcomes (i.e., an ambiguous definition of technical failure); 2) studies including pediatric populations; 3) studies using other elastography modalities (i.e., transient elastography [TE] or point shear-wave elastography); 4) partially overlapping patient cohorts; 5) case reports or series including less than 10 patients; and 6) reviews, guidelines, consensus statements, editorials, letters, comments, or conference abstracts.
Literature search and study selection were performed by one reviewer and double checked by other two reviewers.
Data Extraction
Data pertaining to the following parameters were extracted using a standardized form: 1) study characteristics: authors, institution, duration of patient recruitment, year of publication, and study design (prospective vs. retrospective); 2) patient characteristics: number of patients, male-to-female ratio, mean age, age range, and etiology; 3) technical characteristics of 2D-SWE: device, manufacturer, transducer, measurement number, representative value (mean or median), and number of observers; and 4) study outcomes: proportion of technical failure, proportion of unreliable measurements, and intraclass correlation coefficient (ICC) for interobserver and intraobserver reliability, if any. Additionally, possible factors influencing technical failure or unreliable measurements in each eligible study were evaluated.
The data extraction was performed by two reviewers independently. Any disagreements were resolved with a 3rd reviewer. There was no major controversial issue.
Quality Assessment
The methodological quality of the selected studies was assessed by one reviewers using tailored questionnaires and criteria provided by the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) (43).
Data Synthesis and Analysis
This meta-analysis assessed four main indices: 1) pooled proportion of technical failure; 2) pooled proportion of unreliable measurements; 3) pooled ICC for interobserver reliability; and 4) pooled ICC for intraobserver reliability. If the indices were obtained by two or more observers, especially for the evaluation of intraobserver reliability, representative data (i.e., mean values of all observers' outcomes) were chosen for analysis. Otherwise, data from the observer with the highest value were used.
The pooled proportions of technical failure and unreliable measurements were calculated using meta-analytic pooling via the inverse variance method for calculating weights (44, 45, 46). Random-effects meta-analysis of single proportions was used to obtain an overall proportion. Logit transformation of proportion was performed. The Clopper-Pearson interval for individual studies was used to obtain the confidence intervals (CIs), and a continuity correction of 0.5 was performed in studies with zero cell frequencies. Heterogeneity among studies was determined using 1) Cochran's Q-test for summary estimates with p < 0.05 indicating heterogeneity and 2) the Higgins inconsistency index (I2), which indicates the percentage of variance in a meta-analysis (a rough guide to interpretation: 0–40%, heterogeneity might not be important; 30–60%, moderate heterogeneity may be present; 50–90% substantial heterogeneity may be present; and 75–100%, considerable heterogeneity may be present) (47, 48). Publication bias was assessed using funnel plots visually and Egger's test with p < 0.10 indicating significant bias (49). Publicationbias-adjusted pooled estimate was also calculated using the trim-and-fill method (50). A sensitivity analysis was conducted using a leave-one-out analysis to identify outliers and evaluate the influence of a single study. Moreover, subgroup analyses were performed on the following covariates: 1) measurement numbers (≤ 3 vs. > 3) (7); 2) manufacturer; and 3) etiology (chronic liver disease vs. liver cirrhosis). Specifically, from some of the eligible studies that included both healthy and diseased cohorts (25, 29, 36, 38), we extracted more detailed outcomes of patients with chronic liver disease or liver cirrhosis. Thus, for subgroup analyses on different etiologies (chronic liver disease vs. liver cirrhosis), we also included these subgroup data.
To calculate the pooled ICC for interobserver and intraobserver reliability, we used the Hedges-Olkin method with Fisher's Z transformation of the correlation coefficient (51). With this method, the ICC was converted to Z transforms; thereafter, a mean transformed correlation weighted by sample size was calculated. Once a 95% CI was obtained for the pooled Z score, it was transformed back to a 95% CI for the pooled ICC with both fixed- and random-effects models. The value of ICC can be interpreted as follows: < 0.50, poor; 0.50–0.74, moderate; 0.75–0.89, good; and 0.90–1.00, excellent reliability (52). Heterogeneity and publication bias were also assessed in a similar manner to the pooled proportion of technical failure and unreliable measurement.
All statistical analyses were performed by two reviewers (with 2 and 6 years of experience, respectively, in performing systematic reviews and meta-analyses) using the “metafor” and “meta” packages in R software version 3.5.1 (R Foundation for Statistical Computing).
RESULTS
Literature Search and Quality Assessment
Figure 1 illustrates the flow of literature screening and selection. Finally, 34 articles were included in our systematic review and meta-analysis (8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41). All studies satisfied more than half the tailored questionnaires of QUADAS-2 tool (Supplementary Materials in the online-only Data Supplement).
Fig. 1
Flow diagram of study selection.
Characteristics of the Included Studies
The detailed characteristics of the included studies are summarized in Tables 1 and 2. Twenty-eight of the 34 studies were prospective (8, 9, 10, 11, 12, 13, 14, 15, 17, 18, 19, 20, 21, 22, 23, 24, 26, 27, 28, 29, 31, 33, 34, 36, 37, 39, 40, 41) and four were retrospective (16, 30, 32, 38). The mean ages of subjects in the included studies ranged from 27 to 60 years old. The study populations ranged from healthy cohorts to patients with chronic liver disease/liver cirrhosis from various causes.
Table 1
Demographic Characteristics of Included Studies
Table 2
Technical Characteristics of Included Studies
The US device used in 28 studies was Aixplorer (Supersonic Imagine, Strasbourg, France) (9, 10, 11, 12, 13, 15, 16, 17, 18, 19, 20, 23, 25, 26, 27, 28, 29, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41). Either LOGIQ E9 (GE Healthcare, Chicago, IL, USA) (8, 14, 22, 30) or Aplio 500 (Canon Medical Systems, Otawara, Japan) (21, 24) was used in the remaining 6 studies.
Regarding the methods of liver-stiffness measurements, 23 studies performed more than three measurements (8, 10, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 24, 25, 30, 31, 35, 36, 37, 38, 39, 40, 41), whereas 9 studies performed three (9, 11, 13, 23, 26, 29, 32, 33) or less (28) measurements. Thirteen studies used “mean” as a representative value of liver stiffness (9, 10, 13, 14, 15, 16, 19, 21, 23, 25, 26, 33, 40), 17 studies used “median” (8, 11, 12, 17, 18, 20, 22, 24, 29, 31, 32, 34, 36, 37, 38, 39, 41), and 2 studies used both values (30, 35).
Technical Failure
We obtained the proportion of technical failure of 2D-SWE in 20 studies including 6196 patients (9, 10, 11, 12, 13, 16, 22, 24, 25, 26, 27, 28, 29, 32, 34, 36, 38, 39, 40, 41). Under the random-effects model, the pooled proportion of technical failure was 2.3% (95% CI, 1.3–3.9%) (Fig. 2). Significant heterogeneity was noted in Cochran's Q-test (p < 0.01) and Higgins I2 (90%). The funnel plot (Supplementary Fig. 1A in the online-only Data Supplement) and Egger's test (p < 0.01) revealed substantial publication bias. After using the trim-and-fill method (Supplementary Fig. 1B in the online-only Data Supplement), the publication-bias-adjusted pooled estimate was 2.8% (95% CI, 1.7–4.7%), suggesting the robustness of the result against any publication bias. No outlier was found in the sensitivity analysis.
Fig. 2
Forest plots of proportions of technical failure.
CI = confidence interval, F = fixed, R = random.
The results of subgroup analyses for the proportion of technical failure are summarized in Table 3. No significant difference in technical failure proportion was observed between the studies with three or less measurements and those with more than three measurements. Ten studies originally targeted patients with chronic liver disease (9, 10, 11, 12, 13, 16, 27, 28, 40, 41). Additionally, we could extract separate data on patients with chronic liver disease from 4 studies (25, 29, 36, 38). Therefore, we could recalculate the pooled proportion of technical failure in patients with chronic liver disease from 14 studies (2.4%; 95% CI, 1.2–4.8%) (9, 10, 11, 12, 13, 16, 25, 27, 28, 29, 36, 38, 40, 41). Likewise, we recalculated the pooled proportion in patients with liver cirrhosis from 3 studies (6.8%; 95% CI, 2.5–17.0%) (10, 13, 38).
Table 3
Subgroup Analyses for Technical Failure
Unreliable Measurements
From 20 studies including 6961 patients (8, 9, 11, 13, 17, 18, 20, 21, 22, 24, 27, 28, 29, 31, 33, 35, 37, 38, 40, 41), the pooled proportion of unreliable measurement was 7.5% (95% CI, 4.7–11.7%) (Fig. 3). The definition of unreliable measurements varied across the studies (Table 4). Significant heterogeneity was found in Cochran's Q-test (p < 0.01) and Higgins I2 (96%). The funnel plot and Egger's test revealed no significant publication bias (p = 0.19) (Supplementary Fig. 2 in the online-only Data Supplement). One study was an outlier in the sensitivity analysis (13), but the summary proportion was still robust (6.8%; 95% CI, 5.0–9.3%) after removing it.
Fig. 3
Forest plots of proportions of unreliable measurements.
Table 4
Definition of Reliable Measurements in Eligible Studies
The subgroup analyses for the proportion of unreliable measurements are summarized in Table 5. Notably, studies conducting more than three measurements had fewer unreliable measurements than did those conducting three or less measurements. From 12 studies (9, 11, 13, 17, 21, 27, 28, 29, 35, 38, 40, 41), including 2 (29, 38) enabling the extraction of separate data on patients with chronic liver disease, the pooled proportion in patients with chronic liver disease was 6.3% (95% CI, 3.0–12.9%).
Table 5
Subgroup Analyses for Unreliable Measurements
Interobserver and Intraobserver Reliability
The interobserver reliability of 2D-SWE was obtained from 12 studies (10, 12, 14, 15, 18, 19, 20, 21, 22, 23, 36, 39). The study of Yoon et al. (39) was excluded because it potentially shared the population with another (20). Moreover, unlike other studies reporting the result using an ICC parameter, the study of Deffieux et al. (12) used Pearson's correlation coefficient (r = 0.87). Finally, we conducted a meta-analysis of 10 studies including 517 patients (10, 14, 15, 18, 19, 20, 21, 22, 23, 36), and the pooled interobserver reliability was 0.87 (95% CI, 0.82–0.90), suggesting good reliability (Fig. 4A). Significant heterogeneity was noted in Cochran's Q-test (p = 0.01) and Higgins I2 (58%). The funnel plot (Supplementary Fig. 3A in the online-only Data Supplement) and Egger's test (p = 0.08) revealed substantial publication bias, but a publication-bias-adjusted pooled estimate suggested good reliability (ICC = 0.77; 95% CI, 0.74–0.79) under the trim-and-fill method (Supplementary Fig. 3B in the online-only Data Supplement). One outlier was present in the sensitivity analysis (20), and the pooled ICC was 0.88 (95% CI, 0.84–0.90) after removing it, thus suggesting the result was robustness.
Fig. 4
Forest plot of interobserver reliability (A) and intraobserver reliability (B).
ICC = intraclass correlation coefficient
We obtained the intraobserver reliability of 2D-SWE from 7 studies including 679 patients (10, 14, 15, 19, 22, 36, 38). The pooled intraobserver reliability was 0.93 (95% CI, 0.89–0.95) using a random-effects model, suggesting excellent reliability (Fig. 4B). Significant heterogeneity was noted (Cochran's Q-test: p < 0.01; Higgins I2 = 80%). The pooled reliability was still robust (0.95; 95% CI, 0.94–0.96) after removing one outlier (36). We could not calculate the publication bias for intraobserver reliability because of the small sample size (< 10 studies).
Influential Factors
Factors influencing technical performance were reported in 16 studies (8, 10, 11, 16, 18, 20, 22, 24, 25, 27, 29, 33, 35, 37, 38, 40) (Table 6). Overall, technical failure and/or unreliable measurement was affected by patient factors, including high body mass index/wide waist circumference/thick intercostal wall suggestive of overweight or obesity, old age, inability to optimally hold breath, severe liver disease and associated complications (e.g., ascites), narrow intercostal space, and long distance between the transducer and liver capsule. Additionally, one study reported that operator experience significantly influenced the measurement reliability of 2D-SWE (28).
Table 6
Factors Influencing Technical Failures and/or Unreliable Measurements
DISCUSSION
Our meta-analysis revealed that the pooled proportions of technical failures and unreliable measurements of 2D-SWE were 2.3% and 7.5%, respectively. Moreover, the 2D-SWE measurements showed good to excellent interobserver (ICC = 0.87) and intraobserver (ICC = 0.93) reliability, suggesting the applicability of 2D-SWE for evaluating liver stiffness. Our result also revealed that the technical performance of 2D-SWE is comparable to TE, the most extensively used US elastography, reported to have failure rate of 3.1% and unreliable measurement rate of 15.8% from a study of 13379 examinations (53).
Currently, all 2D-SWE systems enable quality assessment of shear-wave measurements and adjust the display when the quality decreases by dropping the offending pixels and excluding them when calculating Young's modulus (54). Additionally, Aplio 500 shows a display of shear waves travelling, suggesting acceptable quality if consecutive lines are parallel. Other manufacturers also provide an additional vendor-specific approach to quality judgement, including confidence maps in Philips systems and the stability index in the new software version of Aixplorer (7).
However, according to the 2017 European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) guidelines, no agreement exists on the quality criteria for 2D-SWE and unreliable measurements are randomly defined across studies (7). Naturally, the proportion of unreliable measurements were closely influenced by the definition, as shown by Elkrief et al. (13) whose proportion was substantial under the strict definition. Some authors (17, 18, 27, 28) used minimal Young's modulus to identify invalid measurements. The Society of Radiologists in Ultrasound consensus (55) and other studies (8, 11, 21, 22, 37, 38, 40, 41) recommend interquartile ranges/median values below 30% as valid measurements mimicking the TE reliability criteria. To reduce such variability and enable standardization, a collaborative effort by academia and manufacturers is required (56).
Given the significant heterogeneity in technical failure and unreliable measurements among the studies in this meta-analysis, subgroup analyses were conducted to explore potential factors influencing successful and qualified measurements. Notably, the number of measurements significantly affected the unreliable measurements; thus, multiple measurements in the same location are recommended for obtaining reliable liver-stiffness measurements (55). Because no consensus exists on the optimal measurement numbers, the included studies performed various numbers of measurements ranging from 3 to 15 (29, 40, 57, 58). Recently, the 2017 EFSUMB guidelines recommended that three measurements suffice to obtain consistent results for assessing liver fibrosis and portal hypertension (7). However, our subgroup analyses revealed that studies conducting more than three measurements showed fewer unreliable measurements than did those conducting three measurements or less. Nevertheless, no difference was found in technical failure between the two groups. Thus, we suggest that the optimal minimum number of 2D-SWE measurements should be further verified. We believe the composition of the population also affects technical success and reliable measurements as liver disease leads to improper procedures caused by changes in liver volumes, secondary interference by an interposed colon, or other complications (38). Three studies targeting patients with cirrhosis had a higher proportion of technical failure, even though studies and study subsets targeting patients with chronic liver disease showed no difference in technical failure and unreliable measurements with the overall population.
2D-SWE helps select a ROI in a representative area of the liver, and it could be saved and followed over time. This could reduce sampling variability in repeated measurements (59). Indeed, good to excellent interobserver and intraobserver reliability in our meta-analysis supports this advantage of 2D-SWE. Conversely, 2D-SWE requires technical expertise because the operators need to consistently place the points of measurements in the liver. One study reported that intraobserver agreement between measurements on different days drops from 0.84 for experienced examiners to 0.65 for beginners (15). Therefore, measurements should be performed by experienced operators, and beginners are recommended at least 50 supervised measurements (7, 15, 60).
Our study has several limitations. Despite the significant heterogeneity in the meta-analysis, we could not conduct further subgroup analyses for potential factors influencing the results and heterogeneity, especially the impact of overweight or obesity on technical performance. Second, many of the included studies used the Aixplorer system because the other manufacturers only recently released their 2D-SWE devices. However, we included all available studies and our results may be generally applied to all 2D-SWE devices. Third, significant publication bias was observed in the meta-analysis for technical failure and interobserver reliability. After using the trim-and-fill method, however, the outcomes were still robust.
In conclusion, 2D-SWE has good technical performance for assessing liver stiffness, being characterized by high technical success and reliability. Nevertheless, future studies should establish the quality criteria and optimal number of measurements.
Supplementary Materials
The online-only Data Supplement is available with this article at https://doi.org/10.3348/kjr.2018.0812.
SUPPLEMENTARY MATERIALS
Click here to view.(22K, pdf)
Results of Tailored Quality Assessment of Diagnostic Accuracy Studies-2 AssessmentSupplementary Table 1
Funnel plot for technical failure before (A) and after (B) trim-and-fill methods. Filled circle: included studies, Open circle: imputed studies identified through trim-and-fill methods.Supplementary Fig. 1
Funnel plot for unreliable measurements.Supplementary Fig. 2
Funnel plot for interobserver reliability before (A) and after (B) trim-and-fill methods. Filled circle: included studies, Open circle: imputed studies identified through trim-and-fill methods.Supplementary Fig. 3
This study was supported by a grant (No. 2016-719) from the Asan Medical Center, Seoul, Korea and a grant (No. 2017R1A2B3011475) from the National Research Foundation of Korea.
Conflicts of Interest:The authors have no potential conflicts of interest to disclose.
References
-
Lupșor-Platon M, Badea R, Gersak M, Maniu A, Rusu I, Suciu A, et al. Noninvasive assessment of liver diseases using 2D shear wave elastography. J Gastrointestin Liver Dis 2016;25:525–532.
-
-
Dietrich CF, Bamber J, Berzigotti A, Bota S, Cantisani V, Castera L, et al. EFSUMB guidelines and recommendations on the clinical use of liver ultrasound elastography, update 2017 (long version). Ultraschall Med 2017;38:e16–e47.
-
-
Elkrief L, Ronot M, Andrade F, Dioguardi Burgio M, Issoufaly T, Zappa M, et al. Non-invasive evaluation of portal hypertension using shear-wave elastography: analysis of two algorithms combining liver and spleen stiffness in 191 patients with cirrhosis. Aliment Pharmacol Ther 2018;47:621–630.
-
-
Gerber L, Kasper D, Fitting D, Knop V, Vermehren A, Sprinzl K, et al. Assessment of liver fibrosis with 2-D shear wave elastography in comparison to transient elastography and acoustic radiation force impulse imaging in patients with chronic liver disease. Ultrasound Med Biol 2015;41:2350–2359.
-
-
Sporea I, Bota S, Gradinaru-Taşcău O, Sirli R, Popescu A, Jurchiş A. Which are the cut-off values of 2D-shear wave elastography (2D-SWE) liver stiffness measurements predicting different stages of liver fibrosis, considering transient elastography (TE) as the reference method? Eur J Radiol 2014;83:e118–e122.
-
-
Higgins J, Green S. Cochrane handbook for systematic reviews of interventions. Version 5.1.0. The Cochrane Collaboration Web site. [Updated March 2011]. [Accessed January 8, 2017].https://handbook-
5- 1.cochrane.org/.
-
-
Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw 2010;36:1–48.
-
-
Castéra L, Foucher J, Bernard PH, Carvalho F, Allaix D, Merrouche W, et al. Pitfalls of liver stiffness measurement: a 5-year prospective study of 13,369 examinations. Hepatology 2010;51:828–835.
-
MeSH Terms
Figures
Tables
Funding Information
-
Asan Institute for Life Sciences, Asan Medical Center
2016-719
-
National Research Foundation of Korea
2017R1A2B3011475