AI Tool Maps Two Distinct Centrosome Defects in Breast Tumours

Overview

• Researchers at the University of Southampton used CenSegNet, an open‑source AI platform, to analyse more than 330,000 centrosomes in tissue from 127 breast cancer patients.
• CenSegNet separated centrosome abnormalities into two distinct defects—excess centrosome number and abnormally enlarged centrosomes—that can occupy different regions within the same tumour.
• Tumours with high levels of enlarged centrosomes were associated with more aggressive features such as higher grade, lymph node involvement and certain genetic changes, and with worse patient survival.
• The team plans to combine CenSegNet outputs with genomic, transcriptomic and proteomic data and to test larger, diverse cohorts to develop centrosome‑based biomarkers for risk stratification and treatment guidance.
• Centrosomes are tiny cell structures that control DNA segregation and have been hard to study in patient tissue; mapping them at single‑cell resolution could reveal tumour heterogeneity and help tailor care if prospective validation succeeds.