Semiautomatic three-dimensional ultrasound renal volume segmentation in pediatric hydronephrosis: interrater agreement and correlation to conventional hydronephrosis grading

Abstract

Background: Two-dimensional (D) assessment of renal volume underestimates the actual value and shows high interobserver variability. Limited data exist on innovative 3-D ultrasound (US) technique for the evaluation of pediatric hydronephrosis.

Objective: To assess the interrater agreement of kidney volume segmentation by 3-D US with a matrix array transducer in children with hydronephrosis and to compare the 3-D metrics to conventional hydronephrosis grading.

Materials and methods: We prospectively acquired 48 renal volumes in 45 patients with hydronephrosis by freehand 3-D US (6-1 MHz volumetric sector array, electronic rotation; median age, 4 years; 1 month to 16 years). Semi-automated kidney segmentation was performed by two independent readers providing volumes for total kidney (renal capsule), dilated collective system, renal parenchyma (renal capsule-collective system) and hydronephrosis index (renal parenchyma/capsule). Interrater agreement was evaluated with Bland-Altman plots, intraclass correlation coefficient (ICC) and Dice similarity coefficients. The maximum calibre of renal pelvis was measured and hydronephroses were morphologically classified grade 1-4.

Results: Interrater agreement for renal capsule, collective system, hydronephrosis index, and renal parenchyma was good to excellent with ICC of 0.94, 0.87, 0.83 and 0.92 respectively (P < 0.001 each). Median Dice was 0.90 (capsule), 0.77 (collective system) and 0.88 (parenchyma). There was a positive correlation between conventional hydronephrosis grading and ultrasonic hydronephrosis index and between renal pelvis diameter and collective system volume (P < 0.001 both).

Conclusion: Semiautomatic 3-D US volumetric analysis has a high degree of interrater agreement and enables reliable assessment of renal parenchymal volume in hydronephrosis. Volumes of the collective system and hydronephrosis index correlate with the extent of hydronephrosis.

Trial registration: Trial registration number, DRKS00022772; date of registration, 07/31/2020.

Keywords: Hydronephrosis; Imaging, three-dimensional; Kidney diseases; Pediatrics; Ultrasonography.

Fig. 1

Workflow of the study. Image parameters were adjusted based on the conventional 2-dimensional (D) B-mode ultrasound. Afterward, the 3-D mode display was used to show live side-by-side-image including a sagittal and axial view of the kidney at the same time. After 3-D volume acquisition the multiplanar reformatted (MPR) view highlights the structure in three orthogonal planes and a volume-rendered image. The segmentation was performed independently by two radiologists on a separate workstation

Fig. 2

Bland-Altmann plot for all renal parenchymal values. Nearly all renal parenchymal volumes were within the limits of agreement as a tolerance range (broken green lines). There is a single outlier below the lower limit of agreement (arrow). SD standard deviation of the differences

Fig. 3

Boxplots of Dice similarity coefficients for total renal volume, collective system, and parenchyma. The highest median values (bold lines) of Dice similarity coefficient (DSC) are found for total renal volume (capsule: median, 0.9) and for the parenchyma (median: 0.88). The collective systems show lower DSCs (median: 0.77) with a relatively wide range (from 0.54 to 0.89). The Dice values for renal parenchyma segmentation are lower than those for capsule segmentation but remain high. Since the delineation of the renal capsule is semi-automatic, its high spatial agreement seems to compensate for the discrepancies in the collective system segmentation. This can also be explained by the fact that the volume of the latter is generally much smaller than that of the former

Fig. 4

Correlation between hydronephrosis indices and clinical grading. Lower values of the hydronephrosis index (renal parenchyma divided by total capsule) are associated with a higher degree of hydronephrosis (P < 0.001)