AI in Mammography

Improving breast cancer screening with artificial intelligence

We’re working with clinicians, patients, and partners to build an AI system for mammography, which could help radiologists detect breast cancer more accurately, quickly, and consistently.

2 million people are diagnosed with breast cancer each year

More than 2 million people are diagnosed with breast cancer each year

Breast cancer is the most common form of cancer globally, and early detection through breast cancer screening can lead to better chances of survival. While screening is critical to improving outcomes, a shortage of specialists around the world means that screening systems are often overburdened, leading to long, anxiety-filled wait times for people awaiting results.

Our research shows that AI can detect breast cancer with the accuracy of a radiologist

The artificial intelligence-powered system integrates into breast cancer screening workflows to help radiologists identify breast cancer earlier and more consistently. Our published research shows that our technology can identify signs of breast cancer as well as trained radiologists.

The AI-system is trained on thousands of de-identified mammograms

The model uses Google’s artificial intelligence technology to learn the complex features in mammograms that are likely to represent signs of cancer. As a result, the system may spot signs of cancer that some specialists might not be able to see. We hope that by combining human and AI abilities, we might be able to improve cancer detection in the future.

Patients, clinicians, and health professionals guide our work

Learn more about Della Ogunleye’s involvement with our Public Involvement Forum, which informs and advises on how we design, test, and implement AI for mammography.

We’re working with partners to test this technology with real patients and clinicians

Blue graphic representing 1 in 5 people=e30

We’re working with Northwestern Medicine to research how the research model can help prioritize high risk cases and shorten time to diagnosis for screened individuals.

Green graphic representing 1 in 6 people=e30

Through the NHS AI Award, we are working with Imperial College London and three NHS trusts to understand whether the research model can act as a “second independent reader” in UK double read screening systems and allow radiologists to focus on high priority cases while improving consistency and quality of screening.

Improving access for those who need it most

Bringing research to reality through partnership

We are working with leading partners like iCAD to accelerate the clinical adoption of our mammography AI technology. iCAD will validate and incorporate our technology into its ProFound Breast Health Suite, making it available to patients around the world.

Types of cancers and outcomes vary by demographic

Breast density varies by race and ethnicity, so ensuring that our datasets are representative of different demographics is crucial in creating an effective and inclusive tool for all, regardless of race or ethnicity.

We are partnering with the Japanese Foundation for Cancer Research (JFCR) at Ariake Hospital to improve the inclusivity of our models, and are continuing to explore how we can make our products more representative.

Photo of Prof. Yusuke Nakamura
Prof. Yusuke Nakamura

Director for Strategic Innovation Promotion Program “AI Hospital” initiative, Japanese Cabinet Office and Advisor of The Cancer Institute of JFCR

Breast cancer is a cancer that has a very high probability of being cured if found at an early stage. By using AI technology for cancer screening, we can maintain the accuracy of diagnosis and reduce the burden on radiologists, even if many people undergo medical checkup.