AI reveals hidden patterns inside breast cancer

Southampton researchers have developed a powerful AI tool that reveals previously invisible patterns inside breast cancers.

CenSegNet is an open-source AI platform created by researchers at the University of Southampton. It can analyse hundreds of thousands of cells in tumour samples with unprecedented speed and precision.

Working with University Hospital Southampton (UHS), the team have now used this tool to analyse samples from patients with breast cancer.

This has revealed new insights into how tiny structures in a cell called centrosomes change as the tumours grow, spread and evolve.

The technology could help clinicians identify high-risk patients. It could also enable them to predict how breast cancer might progress and deliver more targeted treatments. The results have been published in Nature Communications.

What are centrosomes?

Centrosomes act as the cell's organising hubs. They help cells divide correctly and maintain their structure.

When centrosomes become abnormal, cells can accumulate genetic errors. This is a key trait of cancer.

CenSegNet allows scientists to map centrosome abnormalities across entire tumours. It also means they can identify how these defects vary between different parts of the same cancer.

Dr Salah Elias, at the University's School of Biological Sciences and Institute for Life Sciences, said:

“For more than a century, centrosome abnormalities have been recognised as a hallmark of cancer. But studying them in patient tissues has been extremely challenging.

“CenSegNet allows us to analyse these defects at single-cell resolution across entire tumours and uncover patterns that were previously impossible to see."

Revealing new insights

The team collected tissue samples from 127 breast cancer patients treated at UHS. They used these to analyse more than 330,000 centrosomes across 911 tumour specimens.

The AI uncovered two distinct forms of centrosome abnormality. These had previously been considered part of the same process.

One form involved cells acquiring too many centrosomes. The other involved abnormally enlarged centrosomes.

The researchers found that these defects behave independently. They also found they can occupy different regions of a tumour.

The study also revealed links between centrosome abnormalities and clinical disease features.

Tumours with high levels of enlarged centrosomes were associated with more aggressive characteristics. These included higher tumour grade, lymph node involvement and certain genetic alterations.

Patients whose tumours contained fewer enlarged centrosomes in the tumour core tended to have better overall survival.

Dr Elias added: “Rather than viewing centrosome abnormalities as a single phenomenon, our study shows that they have distinct biological states with different spatial distributions and clinical associations.

“Specific combinations of defects may influence how a tumour grows, invades surrounding tissues and responds to treatment. This opens the door to developing new biomarkers and, ultimately, more personalised treatment strategies.”

Future uses of the technology

The technology is not yet ready for routine clinical use. Even so, the findings point towards several future applications.

First, maps of centrosome abnormalities could help clinicians identify patients with aggressive tumours. This could help refine cancer risk beyond current approaches.

Second, the study highlights new opportunities for precision oncology. Several drugs already in development target proteins that control centrosome function. Clinicians may one day be able to use the tool to identify tumours with specific centrosome defects. They could then match patients to treatments that target these.

Third, the work could help reveal why different parts of the same tumour behave differently. This could potentially help researchers predict tumour progression, metastasis and treatment resistance.

Encouraging use worldwide

CenSegNet is freely available as open-source software. The researchers hope it will be adopted by cancer scientists and pathologists globally.

The team has already demonstrated that the technology can be applied to tissues beyond breast cancer. These include kidney, colon and appendix samples.

The researchers believe this could help establish a new field where AI is used to track disease. This would analyse the behaviour of individual cellular structures across entire tissues.

The team now plans to combine CenSegNet with genomic, transcriptomic and proteomic data.

Their aim is to determine whether centrosome-based biomarkers can help guide treatment decisions. They hope this will improve outcomes for patients with cancer.