Volume Assessment of Pediatric Liver Using 3D Ultrasound Dataset

Abstract

Purpose: To evaluate the reliability and validity of three-dimensional (3D) ultrasound (US) volumetric measurements of the liver in children. Methods: Sixteen consecutive children (mean age: 5.75 years, range: 1-10 years of age) who underwent both 3D liver US and CT examinations within the same week comprised our study group. 3D US data were acquired by one experienced radiologist using an US machine equipped with a position sensing system. The reference liver volume was calculated from the CT scan using the voxel-count method. Two radiologists who were blinded to the reference CT liver volume measured the liver volume two times from the 3D US data. For evaluation of validity, the mean value of two US liver volume measurements for each patient were compared with the reference CT liver volume. The reliability of the 3D US liver volume measurements was assessed by calculating the intraclass correlation coefficient (ICC) and Spearmans correlation coefficient. Bland-Altman plots were created. Results: The mean and standard deviation of the liver volumes with 3D US and CT was 658.32 ± 222.92 ml and 665.82 ± 220.49 ml, respectively. The mean absolute volume difference between the 3D US and reference CT liver volumes was -7.5 ml (range, -173.01 ml to 143.63 ml). The ICCs for intraobserver reliability were 0.997 (95% confidence interval, 0.897-0.987) for observer A and 0.997 (95% confidence interval, 0.936-0.992) for observer B. The interobserver ICC was 0.964. Regarding the validity of the 3D US measurement, the ICC between the 3D US and CT was 0.922 (0.792–0.972). The Bland-Altman plots showed good repeatability. For reproducibility, the measured 3D US liver volume by observer A was larger than that obtained by observer B on Bland-Altman plot. The range of limit of agreement on the Bland-Altman plot for validity was -178.3 mL to 163.3 mL Conclusion: Liver volume measurement in children with the 3D US system showed good repeatability. For clinical applications, an observer consensus and detailed instructions for measurement should be implemented.