Atlas 2: Foundation models built for clinical pathology
Meet Atlas 2 — a new family of pathology vision foundation models designed to close the gap to clinical deployment.
Why it matters: Previous models often forced a trade-off between accuracy, robustness, and compute cost. Atlas 2, Atlas 2-B, and Atlas 2-S aim to balance all three.
- Trained on 5.5 million histopathology whole-slide images from Charité – Universitätsmedizin Berlin, LMU Munich, and Mayo Clinic — the largest dataset of its kind.
- Tested across 80 public benchmarks with state-of-the-art prediction performance, robustness, and resource efficiency.
- Multiple sizes to fit real-world constraints, from high-performance to compute-friendly.
Big picture: More reliable and efficient pathology AI could help labs scale quality control, triage, and research workflows — moving trustworthy AI closer to the clinic.
Read the paper: https://arxiv.org/abs/2601.05148v1
Paper: https://arxiv.org/abs/2601.05148v1
Register: https://www.AiFeta.com
#AI #DigitalPathology #MedicalImaging #FoundationModels #MachineLearning #Atlas2 #HealthTech
