UroVision: a new artificial intelligence software for functional magnetic resonance urography evaluation in children with congenital anomalies of the kidney and urinary tract.

OBJECTIVES

This study compared UroVision, a novel AI-based tool for automated three-dimensional (3D) segmentation and magnetic resonance urography (MRU) functional analysis, with the validated ImageJ MRU software, focusing on split renal function (SRF) and drainage curve classification in children with congenital anomalies of the kidney and urinary tract (CAKUT).

MATERIALS AND

METHODS

MRU examinations in children with CAKUT were retrospectively analysed using UroVision and ImageJ MRU. UroVision functional assessment was limited to the two-dimensional slice showing the largest amount of renal parenchyma, matching the ImageJ MRU workflow. Two paediatric radiologists independently performed the evaluations. Agreement was assessed using Pearson correlation, linear regression, Bland-Altman analysis, and intraclass correlation coefficients (ICC). Drainage curve morphology was evaluated by an expert pediatric radiologist, with inter-method agreement assessed using weighted Cohen's κ. The full 3D UroVision analysis was reviewed to aid the interpretation of borderline and discordant cases.

RESULTS

Fifty patients (28 males, 22 females; mean age 2.4 years, range 0-18 years) were included, yielding 100 kidneys for validation. UroVision showed excellent agreement with ImageJ MRU: renal volume correlation was high (r = 0.99), while functional parameters showed moderate correlation (glomerular slope r = 0.63; Patlak coefficient r = 0.57). Bland-Altman analysis showed minimal bias. Inter-reader and intra-reader reproducibility were excellent (ICCs: 0.87-1.00 and 0.89-0.99, respectively). Agreement for drainage curves was similarly high (weighted κ = 0.86). Full 3D analysis enabled reclassification of borderline curves in 16 out of 23 kidneys.

CONCLUSIONS

UroVision provides results comparable to ImageJ MRU, while offering automated analyses and improved reproducibility.

KEY POINTS: Question Can UroVision provide reliable quantitative and qualitative renal and urinary tract functional assessment in children with CAKUT compared with ImageJ MRU? Findings UroVision showed excellent agreement with ImageJ MRU for renal function parameters and drainage curve classification, with high inter- and intra-reader reproducibility. Clinical relevance UroVision uses AI-based automated MRU analysis to deliver reproducible and standardised functional assessment in children with CAKUT, reducing operator dependence and improving workflow efficiency while providing reproducible and standardised functional assessment in clinical practice.