Development of a deep learning model for glomerulus detection and segmentation in kidney histopathology images
- 1The Jikei University School of Medicine
Abstract
BackgroundEstimation of nephron number, an indicator of renal reserve, requires quantitative assessment of glomerular number and size in kidney biopsy specimens. We aimed to develop an automated system for quantitative measurement of glomerular areas from clinical kidney biopsy images, across diverse kidney diseases.
MethodsWe developed a two-stage deep learning system combining Faster R-CNN for glomerular detection in whole-slide images (WSIs) and DeepLabV3+ for segmentation. Of 114 kidney biopsies (228 WSIs), 104 were used for model development and 10 for independent final evaluation. Agreement was assessed using the F1-score for detection, the Dice coefficient for segmentation, and the intraclass correlation coefficient (ICC) for glomerular area measurement. Analysis was performed at 40x total magnification (2.331 m/pixel) and validated against 400x images.
ResultsIn the independent evaluation set, the artificial intelligence (AI) model achieved a mean F1-score of 0.851 for detection, compared with 0.888 among experts. For segmentation, the Dice coefficient was 0.959 {+/-} 0.011 and the per-pair ICC (2,1) for area measurement was 0.989 {+/-} 0.009; both were comparable to inter-expert agreement. Bland-Altman analysis showed a mean bias of -344 m{superscript 2} (95% limits of agreement, -1904 to 1216 m{superscript 2}) between 40x and 400x measurements.
ConclusionThe two-stage model achieved glomerular segmentation and area measurement comparable to expert agreement in independent clinical data, with detection performance approaching inter-expert agreement, and provided reliable measurements at low magnification. This automated approach may facilitate large-scale quantitative kidney pathology and nephron number estimation from clinical kidney biopsies.
Related
- Deep learning based edge segmentation and classification for plasterboard defect detection
- Deep Learning for Real-Time Image Interpretation in Arthroscopy: A Systematic Review of Classification, Segmentation, and Detection Tasks
- Does Deep Learning Vascular Segmentation on CTA Improve Vertebral Artery Dissection Detection?
- Multi-Backbone U-Net segmentation and interpretable hybrid deep learning classification for lung cancer detection in chest CT images
- Inferring clinically relevant molecular subtypes of pancreatic cancer from routine histopathology using deep learning
BibTeX
@misc{tosaki2026development,
title = {{Development of a deep learning model for glomerulus detection and segmentation in kidney histopathology images}},
author = {T. Tosaki and T. Sasaki and Y. Okabayashi and K. Haruhara and M. Okabe and N. Tsuboi and T. Yokoo},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.19.26365514},
url = {https://www.medrxiv.org/content/10.64898/2026.10.19.26365514v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Development of a deep learning model for glomerulus detection and segmentation in kidney histopathology images
AU - T. Tosaki
AU - T. Sasaki
AU - Y. Okabayashi
AU - K. Haruhara
AU - M. Okabe
AU - N. Tsuboi
AU - T. Yokoo
PY - 2026
DA - 2026/10/19
PB - medRxiv
DO - 10.64898/2026.10.19.26365514
UR - https://www.medrxiv.org/content/10.64898/2026.10.19.26365514v1
AB - BackgroundEstimation of nephron number, an indicator of renal reserve, requires quantitative assessment of glomerular number and size in kidney biopsy specimens. We aimed to develop an automated system for quantitative measurement of glomerular areas from clinical kidney biopsy images, across diverse kidney diseases. MethodsWe developed a two-stage deep learning system combining Faster R-CNN for glomerular detection in whole-slide images (WSIs) and DeepLabV3+ for segmentation. Of 114 kidney biopsies (228 WSIs), 104 were used for model development and 10 for independent final evaluation. Agreement was assessed using the F1-score for detection, the Dice coefficient for segmentation, and the intraclass correlation coefficient (ICC) for glomerular area measurement. Analysis was performed at 40x total magnification (2.331 m/pixel) and validated against 400x images. ResultsIn the independent evaluation set, the artificial intelligence (AI) model achieved a mean F1-score of 0.851 for detection, compared with 0.888 among experts. For segmentation, the Dice coefficient was 0.959 {+/-} 0.011 and the per-pair ICC (2,1) for area measurement was 0.989 {+/-} 0.009; both were comparable to inter-expert agreement. Bland-Altman analysis showed a mean bias of -344 m{superscript 2} (95% limits of agreement, -1904 to 1216 m{superscript 2}) between 40x and 400x measurements. ConclusionThe two-stage model achieved glomerular segmentation and area measurement comparable to expert agreement in independent clinical data, with detection performance approaching inter-expert agreement, and provided reliable measurements at low magnification. This automated approach may facilitate large-scale quantitative kidney pathology and nephron number estimation from clinical kidney biopsies.
ER -