Analytics in Action News

Deep Learning Model for Mammograms Predicts Breast Cancer Risk

Researchers developed a deep learning model to identify patients at high risk for breast cancer.

deep learning

Source: Getty Images

By Erin McNemar, MPA

- Deep learning can distinguish between the mammograms of women who will later develop breast cancer and those who will not, according to new research out of the University of Hawaii. Researchers said the findings show the potential of artificial intelligence to act as a second reader for radiologists, reducing unnecessary imaging and associated costs.

Annual mammography is recommended for women to screen for breast cancer starting at the age of 40. Research indicates that screening mammography lowers breast cancer mortality by decreasing the likelihood of cancer advancing undetected.

Mammograms not only assist in detecting cancer but can also predict breast cancer risk by measuring breast density. While denser breasts on mammography are typically associated with a higher risk of cancer, other unknown factors hidden in the mammogram could contribute to risk.

“Conventional methods of breast cancer risk assessment using clinical risk factors haven’t been that effective,” study lead author John A. Shepherd, PhD, said in a press release. “We thought that there was more in the image than just breast density that would be useful for assessing risk.”

In the new study, Shepherd and his colleagues used a data set of over 25,000 digital screening mammograms from 6,369 women. More than 1,600 of the women developed screening-detected breast cancer, and 351 developed interval invasive breast cancer.

The research team trained the deep learning model to detect details in the mammogram linked to increased cancer risks. When they tested the deep learning-based model, it underperformed in assessing the risk factors for interval cancer risk. However, the model outperformed clinical risk factors, including breast density, in determining screening-detected cancer risk.

“The results showed that the extra signal we’re getting with AI provides a better risk estimate for screening-detected cancer,” Shepherd said. “It helped us accomplish our goal of classifying women into low risk or high risk of screening-detected breast cancer.”

These findings could significantly impact clinical practices in which breast density alone guides many management decisions.

Instead of being advised to return next year for another screening, women with a negative mammogram could be sorted into one of three pathways depending on risk: low risk of breast cancer, elevated screening-detected risk, or elevated interval invasive cancer in the next three years, the average follow-up time for the study.

“This would allow us to use a woman’s individual risk to determine how frequently she should be monitored,” Shepherd said. “Lower-risk women might not need to be monitored with mammography as often as those with a high risk of breast cancer.”

Additionally, the deep learning model could assist in supporting decisions regarding additional imaging with MRI and other methods. According to Shepherd, women in the high-risk deep learning group who also have dense breasts and are at a higher risk for interval cancers could benefit most from a monitory strategy including supplemental imaging.

 Along with the other recent research, the study supported the role of artificial intelligence in combination with clinical risk factors in determining breast cancer risk.

“By ranking mammograms in terms of the probability of seeing cancer in the image, AI is going to be a powerful second reading tool to help categorize mammograms,” Shepherd said.

The research team also plans to re-create the study in Native Hawaiian and Pacific Islander women, two underrepresented groups in breast cancer research.