A combined AI model was tested on a group of more than 119,000 women who participated in a breast cancer screening program in Denmark from November 2012 to December 2015. The average patient age was 59 years old.

The researchers employed both a mammography diagnostic platform, Transpara Breast Care, and a texture AI model capable of measuring breast density, which often impacts the outcome of a cancer risk assessments. The algorithms were trained separately using a dataset of more than 39,000 exams. By merging them using a three-layer neural network, the researchers sought to evaluate whether the combination was better at spotting breast lesions that could develop into cancer.

Compared to utilizing the diagnostic and texture models independently, the combined AI model demonstrated enhanced risk assessment for both interval and long-term cancer detection. Interval cancers, which are detected between routine screenings, were also notably identified with increased accuracy by the combined model.

One of the study's most noteworthy findings was their combined AI model's ability to identify high-risk women. Those flagged by the combined model as having the highest 10% combined risk accounted for a significant proportion of interval (44.1%) and long-term (33.7%) cancer cases.

“Current state-of-the-art clinical risk models require multiple tests such as blood work, genetic testing, mammogram and filling out extensive questionnaires, all of which would substantially increase the workload in the screening clinic,” Lauritzen said. “Using our model, risk can be assessed with the same performance as the clinical risk models but within seconds from screening and without introducing overhead in the clinic.” 

While the study showcases promising outcomes, further research and validation will be crucial before widespread adoption of AI-enhanced breast cancer risk assessment can be considered.