New Atlas 2 pathology models achieve SOTA performance on 5.5M images

Researchers have introduced Atlas 2, Atlas 2-B, and Atlas 2-S, a suite of pathology foundation models designed to overcome limitations in performance, robustness, and computational demands for clinical deployment. These models were trained on an unprecedented dataset of 5.5 million histopathology whole slide images sourced from Charité, LMU Munich, and Mayo Clinic. Comprehensive evaluations across eighty public benchmarks demonstrate that Atlas 2 models achieve state-of-the-art prediction performance and resource efficiency. AI

IMPACT These models could significantly improve diagnostic accuracy and efficiency in clinical pathology settings.

RANK_REASON Research paper detailing new foundation models for clinical pathology. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Atlas 2-B
• Atlas 2-S
• Hugging Face
• Ludwig-Maximilians-Universität München
• Mayo Clinic
• Timo Milbich

• paper
• model release
• infra

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