Quantitative evaluation of chronically obstructed kidneys from noncontrast computed tomography based on deep learning

Area of Science:

• Radiology
• Medical Imaging
• Nephrology

Background:

• Chronic kidney obstruction impacts renal morphology and function.
• Accurate quantitative assessment of kidney structure is crucial for evaluating obstruction.
• Noncontrast computed tomography (NCCT) is a common imaging modality.

Purpose of the Study:

• To quantitatively report renal parenchymal volume (RPV), renal sinus volume (RSV), and renal parenchymal density (RPD) in chronically obstructed kidneys using NCCT.
• To evaluate the association between these quantitative parameters and split glomerular filtration rate (sGFR).

Main Methods:

• Retrospective analysis of 304 NCCT scans from patients with urinary obstruction.
• Development and validation of a 3D U-Net model for parenchyma and sinus segmentation.
• Calculation of RPV, RSV, and RPD using the model, followed by multivariate analysis with sGFR.

Main Results:

• High segmentation accuracy achieved (Dice values: parenchyma 0.95 ± 0.04, sinus 0.90 ± 0.05).
• Obstructed kidneys showed increased RSV and RPD, and decreased RPV and sGFR compared to nonobstructed kidneys (P < .001).
• RPV, RSV, RPD, and age were significantly correlated with sGFR in chronically obstructed kidneys.

Conclusions:

• NCCT combined with deep learning provides a single-procedure solution for evaluating kidney morphology and function in chronic obstruction.
• The developed model accurately quantifies key renal parameters and their relationship with renal function.