June 7, 2023 – Artificial intelligence (AI) algorithms applied to mammograms can predict a girl’s 5-year breast cancer risk higher than the standard clinical risk model, says a study In radiologya journal of the Radiological Society of North America.
Doctors typically calculate a girl's risk of breast cancer using the Breast Cancer Surveillance Consortium (BCSC) model, which calculates a risk rating based on self-reported information akin to the patient's age, family history of the disease, whether the girl has had children before, and whether she has dense breasts.
But patients could also be missing some information, akin to family history, said lead researcher Vignesh A. Arasu, MD, a scientist and practicing radiologist at Kaiser Permanente Northern California in a Press releaseAI has the advantage of “using a single data source: the mammogram itself,” he said.
For the study, researchers examined the medical records of 324,009 women who received a mammogram through Kaiser Permanente Northern California in 2016. A smaller group of 13,628 women was chosen for the evaluation. Of these patients, 4,584 were diagnosed with cancer inside five years of the mammogram in 2016.
The research team used five AI algorithms to find out risk scores based on the 2016 mammograms. These AI risk scores were in comparison with the BCSC clinical risk assessment. The AI outperformed the BCSC model by way of risk over a five-year period.
“This strong predictive performance over the five-year period suggests that AI is identifying both missed cancers and breast tissue characteristics that help predict future cancer development,” Arasu said. “In mammograms, we can track breast cancer risk. This is the 'black box' of AI.”
According to the study, the AI algorithms were particularly good at predicting whether a patient would develop breast cancer inside a 12 months of a mammogram.
The researchers said that AI combined with traditional models may lead to earlier and more personalized breast cancer diagnoses.