This study developed and validated an artificial intelligence (AI)-assisted diagnostic model for automated glomerular nephritis (GN) diagnosis based on kidney biopsy images, demonstrating high accuracy and stability. Through multicenter data validation, the article highlights the AI model's robust performance in diagnosing four GN types (IgA nephropathy, membranous nephropathy, focal segmental glomerulosclerosis, and minimal change disease), providing an essential tool to enhance pathologists' work efficiency and diagnostic consistency.
Literature Overview
The article "Artificial intelligence-assisted diagnosis of glomerular nephritis using a pathological image analysis approach: a multicentre model development and validation study", published in eClinicalMedicine, reviews the development and validation process of an AI-assisted diagnostic model for glomerular nephritis. Through multicenter research utilizing 106,988 glomerular microscopic images, the study constructed an integrated model comprising three components: glomerular localization, lesion feature extraction, and patient-level classification. The overall performance evaluation metric, F1-score, exceeded 85%. The research further explored a dual-branch AI model combining light microscopy and immunofluorescence to enhance diagnostic precision.
Background Knowledge
Glomerular nephritis (GN) represents a group of diseases characterized by glomerular inflammation, serving as a primary cause of end-stage kidney disease, particularly in developing countries. Currently, GN diagnosis relies on histopathological analysis of kidney biopsies, a process requiring experienced pathologists to manually identify glomerular lesions. However, this approach suffers from high subjectivity, time consumption, and poor reproducibility. While AI has demonstrated exceptional capabilities in digital pathology analysis for cancers like breast and lung cancer, a unified and efficient framework for AI-assisted GN diagnosis remains lacking. This study pioneers AI application in GN pathology diagnosis by developing an automated, reproducible diagnostic tool that addresses manual slide-review limitations. Although previous research has explored AI's potential in glomerular segmentation and lesion detection, large-scale multicenter data validation remains absent. Through training on over 100,000 images, this study validated the AI model's stability in real-world datasets while identifying its limitations in race-specific adaptation, disease subtypes, and multi-staining methods, providing directions for future research.
Research Methods and Experiments
This study employed a multicenter design, collecting pathological image data from 6,682 kidney biopsy patients. Among these, 1,235 cases were used for training, 312 for internal validation, and 2,483 and 2,652 cases served as two external validation cohorts. All images underwent quality control and standardization, retaining only Periodic Acid-Silver Methenamine (PASM)-stained light microscopy images. The AI model comprises three core modules: the glomerular localization module (GloSNet) for image segmentation, a feature fusion module for extracting and integrating glomerular lesion features, and a patient-level classification module for final diagnosis. Model evaluation metrics included F1-score, precision, recall, and accuracy. Additionally, the study explored a dual-branch model incorporating immunofluorescence (IF) images to further improve diagnostic accuracy.
Key Conclusions and Perspectives
Research Significance and Prospects
This study represents the first demonstration of AI's high accuracy in glomerular nephritis diagnosis, offering pathologists an objective, reproducible auxiliary tool. Future research should expand to more races and GN subtypes while optimizing model interpretability to improve clinical applicability. The team recommends integrating multi-staining methods and multi-omics data to enhance generalization capabilities for broader application across global medical centers.
Conclusion
This study successfully developed an AI-assisted diagnostic model for glomerular nephritis based on kidney biopsy light microscopy images, enabling efficient classification of four GN types (IgAN, MN, FSGS, MCD). The model demonstrated excellent performance in both external validation cohorts, achieving F1-scores exceeding 83%. Visual analysis confirmed the AI model's ability to identify GN-specific pathological features. While current limitations exist in race-specific adaptation and disease subtypes, this tool significantly reduces pathologists' workload and improves diagnostic consistency. Future development directions include expanding to more disease subtypes, integrating multi-staining techniques, enhancing model explainability, and exploring real-time diagnostic and telemedicine applications.
