A retrospective study compared UroVision, a novel AI-based tool for automated 3D segmentation and magnetic resonance urography (MRU) functional analysis, with the validated ImageJ MRU software. UroVision had comparable results to ImageJ MRU, offering automated analyses and improved reproducibility.
Researchers identified patients with known congenital anomalies of the kidney and urinary tract (CAKUT) who underwent at least one MRU examination.
The study included 50 patients. Two paediatric radiologists independently evaluated the examinations using UroVision for automated 3D segmentation and functional analysis, and the established ImageJ MRU software.
For validation, UroVision’s functional assessment was restricted to 2D slice containing the largest amount of renal parenchyma, matching the ImageJ MRU workflow.
Researchers compared renal volume, functional parameters and drainage curve classifications between the two methods. Agreement was assessed using correlation and regression analyses, Bland–Altman analysis and intraclass correlation coefficients.
Drainage curve classification agreement was evaluated using weighted Cohen’s kappa.
UroVision showed excellent agreement with ImageJ magnetic resonance urography software for kidney volume, with a strong correlation. Correlations were moderate for the glomerular slope and Patlak coefficient, which were used to assess renal function.
Bland–Altman analysis showed that UroVision slightly overestimated kidney volume, with a mean bias of 6.92 mL. Bias was minimal for the glomerular slope and Patlak coefficient, at −0.018 and −0.0001, respectively, with no evidence of systematic bias for these functional measurements.
Researchers found that UroVision demonstrated high reproducibility between readers and when the same reader repeated an assessment.
For drainage curve classification, UroVision classified 66 curves as normal, 14 as borderline, and 20 as showing accumulation. ImageJ MRU classified 53 as normal, 23 as borderline, and 24 as showing accumulation.
Researchers concluded that UroVision integrates 3D segmentation, reproducible functional metrics, and PACS compatibility, streamlining renal function assessment and standardising reporting. Future work incorporating multicentre validation, refined models for anatomically complex cases, and further evaluation of the added value of 3D analysis for functional metrics could support its potential implementation in clinical practice.
Reference:
Gnocchi G et al. UroVision: a new artificial intelligence software for functional magnetic resonance urography evaluation in children with congenital anomalies of the kidney and urinary tract. Eur Radiol. 2026. DOI;10.1007/s00330-026-12950-5.
Featured Image: Amonsak on Adobe Stock
Featured Image: Amonsak on Adobe Stock
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
