PodoSighter: A Cloud-Based Tool for Label-Free Podocyte Detection in Kidney Whole-Slide Images

Area of Science:

• Nephrology
• Computational Pathology
• Artificial Intelligence

Background:

• Podocyte depletion is a key factor in progressive kidney disease.
• Current methods for podocyte nucleus detection are manual, time-consuming, and lack precision.

Purpose of the Study:

• To develop an automated tool, PodoSighter, for accurate podocyte nucleus identification and quantification.
• To validate PodoSighter's performance across different species and kidney disease models.

Main Methods:

• Developed PodoSighter, a deep learning-based, cloud-hosted tool for analyzing gigapixel whole-slide images (WSIs).
• Trained the tool using labeled kidney images from mice, rats, and humans.
• Validated PodoSighter on diverse kidney disease models (diabetic kidney disease, crescentic glomerulonephritis) and human biopsies.

Main Results:

• PodoSighter achieved high sensitivity and specificity (up to 0.81/0.91) in detecting podocyte nuclei in mouse, rat, and human periodic acid-Schiff-stained WSIs.
• Extracted podocyte nuclear morphometrics using PodoSighter proved valuable for identifying diseased glomeruli.
• The tool is publicly accessible as cloud-based web application plugins.

Conclusions:

• An automated computational approach for podocyte nucleus detection in standard histology WSIs has been established.
• PodoSighter facilitates podocyte research and holds potential for future clinical applications in kidney disease diagnosis.