A novel automated method for comprehensive renal cast quantification from rat kidney sections using QuPath.

Area of Science:

• Nephrology
• Digital Pathology
• Computational Biology

Background:

• Renal tubular casts are key indicators of kidney injury.
• Accurate quantification of casts is difficult due to variations in morphology and staining.
• Current methods like color thresholding fail to account for diverse cast colors.

Purpose of the Study:

• To develop and validate a novel automated method for quantifying renal tubular casts.
• To overcome limitations of existing quantification techniques, particularly color thresholding.
• To provide a reliable and efficient tool for researchers assessing renal injury.

Main Methods:

• Utilized QuPath, an open-source digital pathology software, with its machine learning pixel classification tool.
• Trained a pixel classifier to identify kidney tissue, various cast colors, and slide backgrounds.
• Validated the method on kidney sections from Dahl rats (high/low salt) and Sprague-Dawley rats (PAN-treated).

Main Results:

• The automated QuPath pixel classifier accurately quantified metachromatic tubular casts.
• This novel method demonstrated superior performance compared to traditional color thresholding.
• The approach proved effective across different rat models and salt conditions.

Conclusions:

• A novel, automated machine learning method for renal tubular cast quantification has been developed.
• This QuPath-based approach offers a comprehensive, efficient, and reliable alternative to existing methods.
• The study provides researchers with a powerful new tool for assessing renal injury through cast quantification.