An artificial intelligence algorithm known as BRAIx risk score, originally developed to detect breast cancer in screening scans, has been adapted into a risk prediction tool capable of estimating a woman’s likelihood of developing the disease within the next four years, according to research published in The Lancet Digital Health.

The tool was developed using mammogram data from nearly 400,000 women and subsequently validated in a cohort of almost 96,000 women in Australia, with findings confirmed in an independent Swedish population of more than 4,500 women. Among women placed in the top 2% by the BRAIx score, nearly one in ten received a breast cancer diagnosis within four years despite having been given the all-clear at screening, a risk level exceeding that associated with inherited BRCA1 or BRCA2 gene mutations.

Current population-based breast screening programmes have reduced breast cancer mortality by approximately 40–50% in women aged 50 to 74, yet they largely apply a uniform approach regardless of individual risk. Traditional risk assessment methods that incorporate genetic data, breast density measurements, and patient questionnaires have had limited uptake in routine clinical practice.

The BRAIx score leverages information already present in mammogram images and outperforms conventional risk factors such as breast density, country of birth, and family history. The authors suggest that AI-based risk stratification could enable more personalised screening pathways, directing closer monitoring towards higher-risk individuals while reducing the frequency of screening for those at lower risk, potentially improving early detection rates without increasing overall costs.

Further studies are required before the tool is considered for integration into routine clinical care.

Published in: The Lancet Digital Health