Mammography is an essential screening and diagnostic tool for the detection of breast cancer and the assessment of breast density. But according to Victoria L. Mango, MD, a breast radiologist at Memorial Sloan Kettering Cancer Center in New York City, AI can help breast imagers and physicians see beyond basic breast density information provided by mammographic images and improve clinical management overall.
Although federal law now requires mammography facilities to include breast density information in reports for patients and their physicians, mammography screening recommendations are dependent on a women’s age and it can be difficult to determine which women with dense breasts will benefit the most from supplemental testing, according to Mango.
“Given recent controversial screening recommendations promoting later and less frequent breast cancer screening for average risk women, it becomes even more vital that we accurately identify from the general population women who are at higher risk for developing breast cancer,” she wrote in a guest editorial published online Feb. 6 in Academic Radiology.
Because of this, and the fact that almost half of women have dense breasts, Mango believes further subclassification of breast cancer risk from mammograms is “imperative” and is where advanced AI algorithms such as convolutional neural networks (CNNs) could be of great benefit.
“Obtaining more individualized risk information from a woman’s mammogram may provide important insight into how to better screen that individual for breast cancer,” she wrote.
CNNs, a type of artificial neural network, can automatically learn to recognize distinctive features from a training set of annotated images, such as mammograms.
Citing a recent study that was successful in using a CNN-based algorithm to assess whether mammographic images can be used to stratify an individual's breast cancer risk through generating heat maps, Mango emphasized the need to provide a more individualized risk assessment beyond a visual breast density assessment.
“Importantly this study highlights the heterogeneity within a density category, with a subset of heterogeneously dense breasts displaying a low risk pattern and some patients with scattered fibroglandular densities demonstrating a high-risk pattern,” she wrote. “This further emphasizes the need to provide a more individualized risk assessment beyond a visual breast density assessment.”
A recent graduate from Dominican University (IL) with a bachelor’s in journalism, Melissa joined TriMed’s Chicago team in 2017 covering all aspects of health imaging. She’s a fan of singing and playing guitar, elephants, a good cup of tea, and her golden retriever Cooper.