This AI tool can find hidden patterns in breast cancer
What's the story
A team of researchers from the University of Southampton has developed an artificial intelligence (AI) tool capable of revealing previously unseen patterns in breast cancer. The technology was used to analyze over 330,000 cells from 127 patients at a local hospital. The findings have major implications for identifying high-risk patients and developing more targeted treatment strategies.
Discovery
Two previously indistinguishable changes in cancer cells
The research led by Dr. Salah Elias from the University of Southampton uncovered two distinct types of changes in cancer cells.
One type involved an overproduction of centrosomes, the cell's organizing centers that ensure proper division and structure maintenance.
The other change was an unusual enlargement of these centrosomes.
Both alterations were previously thought to be the same but are now known to occur in different tumor regions and possibly influence cancer development differently.
Technological advancement
A look at the AI platform
To analyze tumor samples with unprecedented speed and precision, the team developed an AI platform called CenSegNet.
This tool can detect centrosome abnormalities at a single-cell resolution across entire tumors, revealing patterns that were previously impossible to see.
The researchers found that tumors with more enlarged centrosomes were linked to more aggressive disease features such as higher tumor grade and spread to nearby lymph nodes.
Future implications
Potential to improve cancer treatment and management
The researchers believe that CenSegNet could help doctors better understand which cancers are likely to grow, spread, or resist treatment.
It could also assist in developing more personalized treatments by identifying tumors with specific weaknesses that could be targeted with new drugs.
Although the technology isn't ready for routine use in hospitals yet, it has been successfully used on tissue samples from other body parts like the kidney, colon, and appendix.