Related Experiment Video
Updated: Jun 9, 2026
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
A multi-task deep sequential neural network for IgA nephropathy Oxford classification and prognosis prediction.
Sai Pan1, Yibing Fu2, Lai Jiang2
1Department of Nephrology, First Medical Center of Chinese PLA General Hospital, Nephrology Institute of the Chinese People's Liberation Army, National Key Laboratory of Kidney Diseases, National Clinical Research Center for Kidney Diseases, Beijing Key Laboratory of Kidney Disease Research, Beijing, China.
DeepSNN, a novel deep sequential neural network, integrates IgA nephropathy (IgAN) diagnosis and prognosis prediction, achieving performance comparable to senior pathologists. This tool streamlines clinical workflows for IgAN patients.
Area of Science:
- Nephrology
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Deep learning models have advanced IgA nephropathy (IgAN) pathological analysis.
- A clinical challenge persists in integrating multi-label structural identification, Oxford classification, and prognosis prediction for IgAN.
- Current models lack comprehensive integration for multi-faceted IgAN assessment.
Purpose of the Study:
- To develop an integrated deep learning model for IgA nephropathy (IgAN) pathological analysis.
- To combine lesion segmentation, Oxford classification, and prognosis prediction into a single framework.
- To enhance the interpretability and clinical utility of AI in IgAN diagnosis.
Main Methods:
- Developed DeepSNN, a deep sequential neural network, as a multi-task model.
- Trained on multi-center, multi-modal renal pathology datasets.
- Integrated subnets for lesion segmentation, glomerular classification, Oxford MEST-C scoring, and prognosis prediction.
- Validated interpretability via visualization and comparison with pathologist diagnostic patterns using Cohen's Kappa.
Main Results:
- DeepSNN achieved high dice coefficients for lesion identification (0.95 and 0.92) on two datasets.
- Demonstrated strong performance in Oxford classification (Kappa values 0.79-0.87).
- Outperformed junior pathologists and matched senior pathologists' performance.
- Showcased superior prognosis prediction (AUC: 0.810) compared to the IIPT (AUC: 0.742).
- Visualization maps confirmed consistent pathological region identification with human experts.
Conclusions:
- DeepSNN successfully integrates multiple diagnostic tasks in IgA nephropathy (IgAN) with high accuracy.
- The model's performance is comparable to senior pathologists, indicating potential for clinical workflow optimization.
- This innovation addresses key gaps in automated renal pathology, offering a clinically interpretable solution for IgAN management.
