Related Experiment Video
Updated: Jul 13, 2025
09:16
Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis
Published on: June 18, 2020
6.9K
Automated Reference Kidney Histomorphometry using a Panoptic Segmentation Neural Network Correlates to Patient
Brandon Ginley1, Nicholas Lucarelli2, Jarcy Zee3,4
1Department of Pathology and Anatomical Sciences, University at Buffalo - The State University of New York, Buffalo, NY, USA.
Summary
Deep learning automated kidney histomorphometry, revealing significant variations in healthy kidney structure related to age, sex, and creatinine levels. This advances quantitative analysis for future large-scale kidney studies.
Area of Science:
- Nephrology
- Computational Pathology
- Medical Imaging
Background:
- Reference histomorphometric data for healthy human kidneys is scarce due to manual quantitation challenges.
- Understanding kidney structure variations is crucial for diagnosing and managing kidney diseases.
- Automated analysis methods are needed to improve efficiency and accuracy in histomorphometry.
Purpose of the Study:
- To develop and apply deep learning for automated histomorphometric analysis of healthy human kidney tissue.
- To investigate the relationships between kidney histomorphometry and patient demographics (age, sex) and serum creatinine.
- To establish a foundation for large-scale quantitative analysis of kidney structure.
Main Methods:
- A panoptic segmentation neural network was developed to segment kidney structures in digitized periodic acid-Schiff (PAS)-stained sections.
- Segmentation included viable/sclerotic glomeruli, cortical/medullary interstitia, tubules, and arteries/arterioles.
- Morphometric parameters were measured, and regression analysis was used to correlate them with age, sex, and serum creatinine.
Main Results:
- The deep learning model achieved high segmentation accuracy across all kidney compartments.
- Significant variations in nephron size/density, vasculature, and interstitium were observed among healthy individuals.
- Nephron size correlated with serum creatinine; renal vasculature and interstitium showed sex-based differences; glomerulosclerosis increased, and arterial density decreased with age.
Conclusions:
- Automated histomorphometric measurements of kidney tissue are feasible and precise using deep learning.
- Even in minimally pathological kidneys, histomorphometric parameters correlate significantly with patient demographics and serum creatinine.
- This deep learning approach enhances the efficiency and rigor of histomorphometric analysis, supporting future large-scale kidney research.
