June 7, 2023 – Artificial intelligence, or AI, algorithms applied to mammograms do a better job of predicting a woman’s 5-year breast cancer risk than the standard clinical risk model, says a study in Radiology, a journal of the Radiological Society of North America.
Doctors usually calculate a woman’s breast cancer risk using the Breast Cancer Surveillance Consortium (BCSC) model, which relies on self-reported information to come up with a risk score. That score is based on information such as a patient’s age, family history of the disease, if a woman has given birth, and whether she has dense breasts.
But the patients may not have some of the information, such as family history, lead researcher Vignesh A. Arasu, MD, a research scientist and practicing radiologist at Kaiser Permanente Northern California, said in a news release. AI has the advantage of using “a single data source: the mammogram itself,” he said.
In the study, researchers looked at medical data for 324,009 women who got mammograms through Kaiser Permanente Northern California in 2016. A smaller group of 13,628 women was chosen for analysis, of which 4,584 patients were diagnosed with cancer within 5 years of the 2016 mammogram.
The research team used five AI algorithms to come up with risk scores based on the 2016 mammograms. Those AI risk scores were compared to the BCSC clinical risk score. AI outperformed the BCSC model for risk through 5 years.
"This strong predictive performance over the five-year period suggests AI is identifying both missed cancers and breast tissue features that help predict future cancer development,” Arasu said. “Something in mammograms allows us to track breast cancer risk. This is the 'black box' of AI."
The AI algorithms were particularly good at predicting whether a patient would have breast cancer within a year of getting a mammogram, the study showed.
The researchers said AI could result in earlier and more individualized breast cancer diagnoses when used along with traditional models.