AI reveals centrosome patterns in breast cancer

Researchers at the University of Southampton developed CenSegNet, an open-source AI platform that analyzes centrosome abnormalities in breast cancer tumors at single-cell resolution. The study, published in Tata Communications, examined over 330,000 centrosomes from 911 tumor specimens of 127 breast cancer patients treated at University Hospital Southampton NHS Foundation Trust. CenSegNet identified two distinct centrosome abnormalities—excess centrosomes and enlarged centrosomes—which were previously considered a single process. The AI revealed that these defects behave independently and occupy different tumor regions. Tumors with high levels of enlarged centrosomes were more aggressive, associated with higher grade, lymph node involvement, and certain genetic changes, while patients with fewer enlarged centrosomes had better survival. The findings could lead to new biomarkers and personalized treatment strategies. The tool has also been applied to kidney, colon, and appendix samples, and the team plans to integrate genomic, transcriptomic, and proteomic data to guide treatment decisions.