Accurate Kidney Pathological Image Classification Method Based on Deep Learning and Multi-Modal Fusion Method with...

Area of Science:

• Nephrology
• Medical Imaging
• Artificial Intelligence

Background:

• Membranous nephropathy is a leading cause of nephrotic syndrome in adults.
• Diagnosis relies on kidney biopsy pathology, a manual and time-consuming process.
• Current diagnostic methods can exhibit inter-physician variability.

Purpose of the Study:

• To develop and evaluate an AI-powered framework for classifying membranous nephropathy.
• To improve diagnostic accuracy and efficiency by integrating multiple imaging modalities.
• To reduce subjectivity in membranous nephropathy diagnosis.

Main Methods:

• Utilized whole-slide images (WSIs) from light microscopy and immunofluorescence microscopy images.
• Developed a framework with glomerular segmentation, feature extraction, and multi-modal fusion modules.
• Trained a glomerular classifier on segmented glomeruli from both image types.

Main Results:

• The combined multi-modal approach achieved an F1-score of 97.32%.
• Individual modalities achieved F1-scores of 92.76% (light microscopy) and 93.20% (immunofluorescence).
• The integrated framework demonstrated superior performance compared to single-modality analysis.

Conclusions:

• Integrating WSIs and immunofluorescence images effectively enhances membranous nephropathy diagnosis.
• The proposed AI framework offers a more accurate and efficient diagnostic tool.
• Multi-modal image analysis holds significant promise for improving kidney disease diagnostics.