Related Experiment Video
Updated: Sep 15, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.3K
Multimodal deep learning improving the accuracy of pathological diagnoses for membranous nephropathy
Xiuxiu Hu1, Jinyue Yang2, Yiping Li1
1Department of Pathology, School of Medicine, Southeast University, Nanjing, China.
Renal Failure
|July 14, 2025
Summary
This study developed an AI system for diagnosing membranous nephropathy (MN) using multimodal pathology images. The AI system demonstrated high accuracy, assisting pathologists and improving diagnostic efficiency for kidney diseases.
Area of Science:
- Nephrology
- Digital Pathology
- Artificial Intelligence
Background:
- Renal biopsy is the gold standard for diagnosing glomerular diseases like membranous nephropathy (MN).
- Current pathological evaluation faces challenges in accuracy, objectivity, and reproducibility.
- Automated diagnostic tools are needed to assist pathologists.
Purpose of the Study:
- To develop a multimodal pathological diagnosis system to aid in diagnosing MN.
- To improve the accuracy and efficiency of MN diagnosis using AI.
Main Methods:
- Developed three deep-learning models using PASM-stained, immunofluorescence, and electron microscopy images.
- Combined model outputs for comprehensive pathological diagnosis.
- Validated the system against pathologists and on external datasets.
Main Results:
- Achieved 91.74% accuracy for spike identification on PASM images.
- Reached 99.31% F1 score for MN classification on immunofluorescence images.
- Demonstrated superior performance to pathologists in specific tasks and high accuracy on external data.
Conclusions:
- The multimodal AI system assists pathologists in rapid and accurate MN diagnosis.
- This system provides a foundation for developing diagnostic models for other glomerular diseases.
