3D ultrasound volume quantification for pediatric urinary tract dilation: a semi-automated segmentation software inter-rater analysis

Abstract

Objective: We determined the reliability of a three-dimensional (3D) ultrasound (US) segmentation software for evaluating volumetric hydronephrosis index (HI) and renal parenchymal and pelvicalyceal volume in children with urinary tract dilation (UTD).

Material and methods: From 1/2019 to 9/2023, children clinically scheduled for a renal imaging exam to assess UTD at a single center were prospectively enrolled. They underwent a dedicated two-dimensional (2D) and 3D US renal exam. A UTD score was assigned per kidney from the 2D images based on the 2014 consensus classification by an experienced pediatric radiologist. From the 3D dataset, the renal parenchyma and collecting system were independently segmented by three trained raters using a semi-automated software. From this segmentation, the kidney parenchymal and pelvicalyceal volume, dimensions, and volumetric HI values were analyzed using the intraclass correlation coefficient to grade inter-rater reliability.

Results: Forty-eight studies from 47 patients were included (65% male; median age 24 months; IQR 61 months). From these, 46 right and 40 left kidneys were chosen based on image quality. Twenty-nine (33.7%) kidneys had no UTD, 10 (11.6%) had UTD P1, 23 (26.7%) had UTD P2, and 24 (27.9%) had UTD P3. Inter-rater reliability was almost perfect across all parameters, with estimates ranging from 0.85 to 0.95. In sub-analysis of kidneys with UTD P2 and UTD P3, volumetric HI had the lowest inter-rater agreement (0.75 and 0.66, respectively).

Conclusions: Semi-automated 3D US segmentation for kidneys with UTD can reliably assess renal dimensions, parenchymal and collecting system volumes, and volumetric HI among raters.

Keywords: Child; Hydronephrosis; Kidney; Three-dimensional imaging; Ultrasonography.