Measurement of Glomerular Filtration Rate using Quantitative SPECT/CT and Deep-learning-based Kidney Segmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Nephrology

Background:

• Quantitative SPECT/CT offers improved glomerular filtration rate (GFR) measurement over planar scintigraphy.
• Manual segmentation of renal parenchyma in CT images for GFR calculation is time-consuming and labor-intensive.

Purpose of the Study:

• To develop a fully automated GFR quantification method using deep learning for 3D kidney segmentation in CT images.
• To assess the accuracy and efficiency of the automated method compared to manual segmentation.

Main Methods:

• A deep learning model was developed for the automated 3D segmentation of kidney parenchyma in CT scans.
• The automated segmentation was validated against manual segmentation using the Dice similarity coefficient.
• Glomerular filtration rate (GFR) was calculated using both manual and automated segmentation methods.

Main Results:

• The automated kidney segmentation achieved a high Dice similarity coefficient (mean = 0.89) compared to manual segmentation.
• GFR values derived from automated and manual methods showed strong correlation (R² = 0.96) with a minimal absolute difference (2.90%).
• The automated method demonstrated comparable performance in diverse patient groups and identified decreased GFR in symptomatic kidneys.

Conclusions:

• The proposed deep learning approach enables fast and accurate GFR measurement through automated kidney segmentation.
• This automated method significantly reduces the time and labor required for quantitative SPECT/CT analysis.
• The findings support the clinical utility of automated segmentation for reliable GFR assessment.