Related Experiment Video
Updated: Jun 2, 2025
10:37
Comparative Proteomic Analysis of Whole Kidney, Medulla, and Cortical Tubules in Diabetic Pathogenesis of Kidney Injury in Mice
Published on: May 2, 2025
94
Exploring the subtle and novel renal pathological changes in diabetic nephropathy using clustering analysis with deep
Tomohisa Yabe1, Yuko Tsuruyama2, Kazutoshi Nomura1
1Department of Nephrology, Kanazawa Medical University, 1-1 Daigaku, Uchinada, 920-0293, Ishikawa, Japan.
Scientific Reports
|January 15, 2025
Summary
Early diagnosis of diabetic kidney disease is crucial for managing chronic kidney disease (CKD). This study used invariant information clustering (IIC) and AI visualization to identify subtle, early pathological changes in diabetic nephropathy.
Area of Science:
- Nephrology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetic kidney disease (DKD) is a major cause of chronic kidney disease (CKD).
- Early diagnosis of DKD is essential for timely intervention and improved patient outcomes.
- Identifying subtle, early pathological changes in DKD is challenging with traditional methods.
Purpose of the Study:
- To apply invariant information clustering (IIC) to glomerular images for early DKD detection.
- To utilize AI visualization techniques (Grad-CAM and GAN) to identify novel pathological changes in DKD.
- To distinguish between diabetic and non-diabetic kidney disease based on glomerular morphology.
Main Methods:
- Clustering of 13,251 glomerular images from 45 patients using IIC into 10 distinct clusters.
- t-distributed stochastic neighbor embedding (t-SNE) for visualizing cluster separation.
- Gradient-weighted class activation mapping (Grad-CAM) and Cycle-Generative Adversarial Networks (Cycle-GAN) for lesion identification.
Main Results:
- IIC successfully clustered images into diabetic-predominant (Clusters 0, 1, 2) and non-diabetic-predominant (Clusters 8, 9) groups.
- Grad-CAM identified characteristic lesions in the outer glomerular capillaries of diabetic clusters.
- Cycle-GAN revealed smaller glomerular tufts relative to Bowman's space as a key lesion in diabetic nephropathy.
Conclusions:
- AI-driven analysis of glomerular images can identify subtle, early pathological changes in diabetic nephropathy.
- IIC and advanced visualization techniques offer a novel approach for early DKD diagnosis.
- These findings may aid in the earlier detection and management of DKD, potentially reducing CKD prevalence.
