Keywords:
Abdomen, Liver, Ultrasound, Elastography, Observer performance, Diagnostic procedure, Cirrhosis, Tissue characterisation
Authors:
B. Fejér, A. Kucsa, Z. Unger, T. Youn, M. Abonyi, V. Bérczi, P. Kaposi; Budapest/HU
DOI:
10.1594/ecr2018/C-1265
Results
The demographic and physiologic variables were compared with MANOVA test between the groups.
There was no significant difference in average (µ) weight (µ=77 kg,
95% CI = 70 - 85 vs. µ=69 kg,
95% CI = 59 - 80 ) or height (µ=169 cm,
95% CI = 163 - 174 vs. µ=173 cm 95% CI =163 - 183) or position of the measurement box from skin surface (µ=4.37 cm,
95% CI = 4.14 - 4.59 vs. µ=4.05 cm,
95% CI = 3.74 - 4.37) between patients and control subjects respectively.
The median age and BMI were both higher in patient group.
The gender distribution was similar between the groups with almost equal representation of both sexes among patients and controls respectively (Table 1).
Interobserver reliability
When paired stiffness values from the two observers were used in a scatter plot,
the Spearman test found significant linear correlation (Figure 2).
The Shapiro-Wilk test indicated that LS differences are normally distributed (p > 0.2).
The mean of the differences (µ=0.036 m/s,
95% CI=-0.08 - 0.15) was centered on zero,
which did not indicate significant measurement bias (Figure 3).
The intra-class correlation found good agreement between LS values measured by the two examiners (ICC=0.71,
95% CI=0.522-0.833,
p<0.001). Cohen's kappa analysis also revealed good reproducibility of the groups with or without significant fibrosis (Table 2).
Prediction rate of significant fibrosis and cutoff LS values
Both observers recorded significantly (p<0.001) increased LS in patients (observer 1 M = 2.17 m/s,
range =1.19 - 2.86 m/s; observer 2 M = 1.93 m/s,
range = 1.28 - 3.19 m/s) compared to healthy controls (observer 1 M = 1.29m/s,
range = 1.18 - 1.56 m/s; observer 2 M=1.36 m/s,
range = 1.16 - 1.85 m/s).
A ROC analysis was performed for both observers to identify optimal cutoff values for significant (≥ F2) liver fibrosis.
The cutoff LS by the first observer was nearly identical compared to the second observer (1.59 m/s vs.
1.61 m/s).
Both observers could differentiate between subject with or without significant fibrosis (AUROC 95.16% vs.
90.62%) with excellent accuracy (Figure 4).