AI Breakthrough: Detecting High Breast Cancer Risk in Women Given All-Clear
A groundbreaking Australian-made artificial intelligence tool has the potential to revolutionize breast cancer screening, identifying women with a high likelihood of breast cancer who would otherwise go undetected. The AI tool, BRAIx, has been developed using data from nearly half a million mammograms and an independent dataset of around 4,500 women in Sweden.
BRAIx can detect risk factors that humans might miss, providing a personalized risk score between 0 and 99.9. This score predicts a woman's likelihood of developing breast cancer within the next four years. The study, published in the Lancet Digital Health journal, highlights the tool's superior accuracy compared to current risk assessment methods.
One of the most remarkable findings is that one in ten women in the top 2% of risk scores went on to develop breast cancer, even though their mammograms were clear. This discovery challenges the traditional approach of relying solely on mammograms for screening.
Dr. Helen Frazer, the clinical director at St Vincent's BreastScreen, expressed her enthusiasm for the algorithm's ability to identify previously undetectable risk signals. She emphasizes the urgency of the situation, with 90,000 Australian women predicted to die from breast cancer in the next 25 years.
The BRAIx algorithm was developed using data from 2016 to 2017, and it outperformed tools based on age, family history, and breast density score. Dense breast tissue, which appears white on a mammogram, can be challenging to detect, but BRAIx can differentiate between dense tissue and cancer with ease.
Jess Armstrong, a cancer survivor, welcomes the AI tool's potential to reduce fear and anxiety associated with mammograms. She highlights the pressure on hospital systems and the potential for human error, advocating for the AI's ability to provide immediate, accurate results.
Dr. Frazer envisions a future where AI-assisted screening starts at age 40 or younger, aiming for zero deaths from breast cancer. She acknowledges the limitations of population screening, which has reduced breast cancer by up to 50% in women aged 50 to 74, but still follows a 'one-size-fits-all' approach.
The development of BRAIx is a significant step forward, according to breast health researcher Wendy Ingman from Adelaide University. She praises the algorithm's ability to define risk levels and predict breast cancer within four years.
The AI tool's potential to support the radiology workforce, which is facing shortages, is another positive aspect. It provides immediate results and could reduce overall system costs by allowing women at low risk to opt for less frequent screenings.
However, the researchers emphasize that AI should not replace radiologists. Human oversight is crucial, as the AI lacks the contextual understanding that humans possess. A focus group study revealed that women are generally open to AI-assisted screening if humans are involved, believing it to be just as accurate as a radiologist.
The Breast Cancer Network Australia's director of policy and advocacy, Vicki Durston, underscores the importance of keeping screening methods up-to-date with scientific advancements. While AI is not a complete solution, it holds the potential to enhance risk assessment and support earlier detection for all women.
The BRAIx tool is the result of a collaborative effort involving researchers from St Vincent's BreastScreen Melbourne, The University of Adelaide, and The University of Melbourne. It received approximately $5 million from the federal government's Medical Research Future Fund, and the researchers are now planning a smaller prospective study to further assess its effectiveness.
