Pathology Foundation Models Show Promise for Mitotic Figure Detection

Researchers have explored the effectiveness of pathology foundation models (FMs) as encoders for mitotic figure detection, moving beyond their typical use in classification tasks. The study compared several FMs, including H-optimus-0 and Virchow models, against a ResNet50 baseline when integrated with different detection architectures like RetinaNet, Faster R-CNN, and Deformable DETR. Results indicate that H-optimus-0 and Virchow models demonstrated competitive performance, suggesting that FM latent spaces, trained via image-level self-supervision, are suitable for direct mitotic figure detection and may offer improved robustness on out-of-domain datasets. AI

IMPACT Demonstrates the potential of foundation models for specialized detection tasks beyond classification in medical imaging.

RANK_REASON Research paper on foundation models for pathology detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Deformable DETR
• Faster R-CNN
• H-optimus-0
• H-optimus-1
• MIDOG++
• ResNet50
• RetinaNet
• TUPAC16
• UNI2-H
• UNI Global Union
• Virchow
• Virchow2

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →