Baseline characteristics

Of the 131 patients who underwent US examination during the study period, 13 patients were unable to hold their breath, the measurement of 10 patients failed due to obvious rib artifacts, and 3 patients had serious underlying diseases (Fig. 1). Thus, a total of 105 adult volunteers were finally included in this study (69 males and 36 females, mean age 52 years, age range 18–75 years). Participants were divided into the non-SLD Group (33 participants with a grade 0) and the SLD Group (72 participants with a visual grade greater than 0) based on the visual grade of hepatic steatosis. Participants’ demographic information and all measured variables are listed in Tables 1, 2, and 3. Some of the measurement states of the UDFF were not adequately measured during the study. The efficiency rate of the measurements in S5, S6, and S7 and end-inspiration was 98.10%, 84.76%, 93.33%, and 99.05%, respectively.

Fig. 1

Study flowchart. In total, 105 participants were included in this study. Finally, 28 of 105 subjects (26%) underwent the measurement of MRI-PDFF. The illustration gives the US and MRI procedures in a patient with moderate steatosis

Table 1 Baseline characteristics of the 105 participantsTable 2 The UDFF of 105 participants in different measurement statesTable 3 The auto-pSWE values of 105 participantsInter-observer and intra-observer variability of the values

The ICC for intra-observer repeatability and inter-observer of UDFF was 0.96 (95% CI: 0.95–0.98) and 0.94 (95% CI: 0.91–0.96), respectively (Fig. 2A). And the intra-observer and inter-observer ICC of auto-pSWE were 0.82 (95% CI: 0.74–0.87) and 0.80 (95% CI: 0.72–0.86), respectively (Fig. 2B). The ICCs of measurements for UDFF and auto-pSWE are all greater than 0.75, indicating that they have good repeatability. For the UDFF measurement, the intra-observer ICCs in period 1 and period 2 were 0.96 and 0.97, and the inter-observer ICCs in the two periods were 0.93 and 0.94, respectively, whereas for auto-pSWE measurement, the intra-observer ICCs in the two periods were 0.73 and 0.87, and the inter-observer ICCs in period 1 and period 2 were 0.71 and 0.84, respectively. The result of the two periods showed that the ICCs of both techniques are higher in period 2 than in period 1, which indicates that training improves the stability of the measurements. The ICCs of UDFF were greater than that of auto-pSWE in different periods, which indicates that the measurement repeatability of UDFF is better than that of auto-pSWE.

Fig. 2

The inter- and intra-observer ICCs and 95% CIs of measurement for overall and each period. A The measurements of UDFF. B The measurements of auto-pSWE

Differences in UDFF values under different conditions

As the UDFF values showed high inter-observer and intra-observer agreement, we further compared the UDFF values in different measurement states. The results are shown in Fig. 3. The median of UDFF values in different hepatic segments were 12 (V), 10 (VI), 12 (VII), and 13 (VIII), respectively; The median UDFF values in different respiratory states were 12 (EE), 12 (EI), and 12 (FB), respectively; The median of UDFF values in different positions were 12 (SP) and 11 (LP). The median UDFF values in different feeding states were 12 (fasting) and 10 (post-prandial). However, none of these differences were statistically significant (p > 0.05). On the other hand, the Covs in different measurement states are significantly different (p < 0.05). The mean Covs for segments V, VI, VII, and VIII were 9.84%, 10.45%, 9.52%, and 6.27%, respectively; The mean Covs for EE, EI, and FB were 6.96%, 9.45%, and 11.31%, respectively; The mean Covs for SP and LP were 7% and 11.11%; The mean Covs for fasting and post-prandial were 7% and 10.56%. The results showed that the segment VIII, end-expiratory, supine, and fasting groups possessed smaller Covs at different measurement states.

Fig. 3

Differences in the UDFF values under various measurement conditions. A The variation in different liver segments. B The variation in different respiratory states. C The variation in different positions. D The variation in different feeding states. EE, end of expiratory; EI, end of inspiratory; FB, free breathing; SP, supine position; LP, lateral position. The data did not satisfy a normal distribution and homogeneity of variance

Correlation of UDFF values with the hepatic steatosis grades and HRI

The UDFF values and hepatic steatosis grades had a significant correlation by a correlated statistical analysis of Spearman’s Rho. The rank correlation coefficient is 0.702 (p < 0.05) (Fig. 4A). The Person correlation analysis showed that UDFF had a positive correlation with HRI values (R2 = 0.069, p < 0.05) (Fig. 4B).

Fig. 4

Scatter plots show the correlation of the UDFF values with the hepatic steatosis grades and HRI. A Correlation between hepatic steatosis grades and UDFF. B Correlation between HRI and UDFF

Linear regression analysis of factors affecting the UDFF and auto-pSWE values

The factors affecting the UDFF and auto-pSWE values are demonstrated in Fig. 5. We included factors in the multivariate linear regression analysis that were significantly less than 0.05 in the univariate regression analysis. According to the univariate analysis, weight (Coef = 0.34, p < 0.001), body mass index (BMI) (Coef = 6.14, p < 0.001), SCD (Coef = 10.77, p < 0.001), hipline (Coef = 0.64, p < 0.001), waistline (Coef = 9.49, p < 0.001), waist–height ratio (whtr) (Coef = 7.55, p < 0.001), and visceral fat area (VFA) (Coef = 9.97, p < 0.001) were associated with the UDFF value (Fig. 5A). SCD (Coef = 4.53, p = 0.043) and VFA (Coef = 4.35, p = 0.032) were factors significantly associated with the UDFF value according to the multivariate linear regression analysis (Fig. 5A). For auto-pSWE, the simple linear regression analysis showed a significant association between it and sex (Coef = 0.07, p = 0.02), BMI (Coef = −0.06, p = 0.002), waistline (Coef = −0.07, p = 0.02), waist–hip ratio (whr) (Coef = −0.12, p = 0.002), and whtr (Coef = −0.1, p = 0.004) (Fig. 5B). The multiple linear regression showed that the significant predictors were sex (Coef = 0.08, p = 0.01) and BMI (Coef = −0.06, p = 0.02) (Fig. 5B). The linear multivariate regression analysis indicated a significant positive association between SCD, VFA, and UDFF (p < 0.05), on the other hand, sex and BMI were correlated to auto-pSWE, but the correlation is not strong.

Fig. 5

The linear regression analysis of measurement parameters. A The liner regression analysis of UDFF. B The liner regression analysis of auto-pSWE. BMI, body mass index; SCD, skin-liver capsule distance; whr, waist-hip ratio; whtr, waist-height ratio; VFA, visceral fat area

Consistency between US-UDFF and MRI-PDFF

We measured MRI-PDFF in 28 of 105 subjects. Of these, 25 were from the SLD group and 3 were from the non-SLD group. Bland–Altman difference plots were used to assess mean bias and 95% limits of agreement (LOA) between UDFF and MRI PDFF measurements in different liver segments. The mean difference between US-UDFF and MRI-PDFF was small (< 2.59) in the four hepatic segments, as well as between them and the whole liver, and the corresponding 95% limits of concordance were less than 21.96% (Fig. 6). The Bland–Altman analysis showed that the mean differences in prediction errors between UDFF and PDFF for liver segments V, VI, VII, VIII were 0.57, 1.69, 1.35, and −0.24, respectively. The result demonstrated that hepatic segment VIII had the lowest mean difference bias compared to other segments. In addition, the concordance analyses showed that the mean difference in prediction error between UDFF for liver segments V, VI, VII, and VIII and PDFF for the whole liver was −0.56, −2.59, −1.10, and 0.73, respectively. The above results suggested both segments V and VIII possessed lower mean difference bias, but segment VIII had smaller concordance boundaries, which may imply that it possesses more stable concordance.

Fig. 6

Bland–Altman plot of the difference between UDFF and MRI-PDFF versus the average of the measures. Solid lines demonstrate the mean difference. The dotted lines represent 95% LOA. A The difference between UDFF and PDFF for S5. B The difference between UDFF and PDFF for S6. C The difference between UDFF and PDFF for S7. D The difference between UDFF and PDFF for S8. E The difference between UDFF of S5 and PDFF of whole liver. F The difference between UDFF of S6 and PDFF of whole liver. G The difference between UDFF of S7 and PDFF of whole liver. H The difference between UDFF of S8 and PDFF of whole liver