Artificial intelligence-powered renal pathology for glomerular disease classification: a systematic review and meta-analysis.
Min Tong, Yingying Jiang, Rongxin Zhu et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the performance of artificial intelligence (AI) models in classifying glomerular diseases using pathology images. The analysis included 15 studies with 39,536 validation sample units. AI demonstrated high performance for membranous nephropathy (sensitivity 0.96, accuracy 0.96), IgA nephropathy (sensitivity 0.92, accuracy 0.94), and minimal change disease (sensitivity 0.92, accuracy 0.96), but evidence comparing AI to pathologists was limited.
Key findings
- AI achieved a sensitivity of 0.96 and accuracy of 0.96 for membranous nephropathy.
- For IgA nephropathy, AI showed a sensitivity of 0.92 and accuracy of 0.94.
- AI demonstrated a sensitivity of 0.92 and accuracy of 0.96 for minimal change disease.
- AI outperformed senior pathologists in accuracy for IgAN, MN, and MCD, though evidence was sparse.
Why it matters
The findings suggest that AI models can significantly enhance the classification of glomerular diseases, potentially improving diagnostic accuracy and efficiency. However, the current evidence does not definitively establish AI's superiority over experienced pathologists, highlighting the need for further validation.
What to keep in mind
Most included studies were retrospective, and there was substantial heterogeneity across datasets and methodologies.