Scientists have known for over 100 years that centrosome abnormalities are linked to cancer, but these tiny structures were nearly impossible to analyze at tumor scale. A new AI system analyzed 330,000 centrosomes from breast cancer patients and found that what researchers had treated as a single type of anomaly is actually two different processes — one of which tracks with more aggressive tumors and worse patient survival. The discovery shows AI’s ability to measure patterns in medical data that were previously effectively invisible.

Related: Frontier Models Outperform Specialized Clinical AI in Head-to-Head Study, General-Purpose LLMs Outperform Specialised Clinical AI Tools, Nature Medicine Study Finds

Independent: AI could help doctors diagnose aggressive breast cancer cases