Atlas 2 -- Foundation models for clinical deployment
Maximilian Alber, Timo Milbich, Alexandra Carpen-Amarie, Stephan Tietz, Jonas Dippel, Lukas Muttenthaler, Beatriz Perez Cancer, Alessandro Benetti, Panos Korfiatis, Elias Eulig, J\'er\^ome L\"uscher, Jiasen Wu, Sayed Abid Hashimi, Gabriel Dernbach, Simon Schallenberg, Neelay Shah, Moritz Kr\"ugener, Aniruddh Jammoria, Jake Matras, Patrick Duffy, Matt Redlon, Philipp Jurmeister, David Horst, Lukas Ruff, Klaus-Robert M\"uller, Frederick Klauschen, Andrew Norgan
Why It Matters
What makes this one worth your time
This work is significant for AI researchers and engineers focused on healthcare, as it addresses key limitations in deploying pathology models clinically, potentially leading to more reliable and efficient diagnostic tools.
Atlas 2 models enhance computational pathology with improved performance and efficiency.
Summary
The paper introduces Atlas 2, Atlas 2-B, and Atlas 2-S, three pathology vision foundation models designed to improve prediction performance, robustness, and resource efficiency in computational pathology. These models were evaluated across eighty public benchmarks and trained on a large dataset of 5.5 million histopathology whole slide images from three medical institutions.
Key contributions
- Development of three new pathology vision foundation models: Atlas 2, Atlas 2-B, and Atlas 2-S.
- Demonstrated state-of-the-art performance and robustness across a wide range of benchmarks.
- Addressed computational efficiency, facilitating potential clinical deployment.
Notable insights
- The models were trained on an unprecedentedly large dataset of 5.5 million histopathology images.
- The evaluation was comprehensive, covering eighty public benchmarks, which suggests thorough validation.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2601.05148v2 Announce Type: replace-cross Abstract: Pathology foundation models substantially advanced the possibilities in computational pathology --- yet tradeoffs in terms of performance, robustness, and computational requirements remained, which limited their clinical deployment. In this report, we present Atlas 2, Atlas 2-B, and Atlas 2-S, three pathology vision foundation models which bridge these shortcomings by showing state-of-the-art prediction performance, robustness, and resource efficiency in a comprehensive evaluation across eighty public benchmarks. Our models were trained on the largest pathology foundation model dataset to date comprising 5.5 million histopathology whole slide images, collected from three medical institutions Charit\'e - Universit\"atsmedizin Berlin, LMU Munich, and Mayo Clinic.