A new, simpler AI model for predicting breast cancer may be able to predict future breast cancer risk in patients up to five years in advance, according to a study by researchers at Duke University in Durham, North Carolina.
Scientists there developed the new model, known as AsymMirai, which they claim is much easier to use than the previous model known as Mirai, as reported in a news release.
The previous model was a “black box — very large and complex neural network similar in construction to ChatGPT — and no one knew how it made its decisions,” explained the study’s lead author, Jon Donnelly, a Ph.D. student in the Department of Computer Science at Duke University. “We developed an interpretable AI method that allows us to predict breast cancer from mammograms 1 to 5 years in advance. AsymMirai is much simpler and much easier to understand than Mirai.”
Researchers from the Department of Computer Science and the Department of Radiology compared 210,067 mammograms from 81,824 patients in the Emory Breast Imaging Dataset gathered over an eight-year period between January 2013 and December 2020, using both models.
“Previously, differences between the left and right breast tissue were used only to help detect cancer, not to predict it in advance,” said Donnelly in the news release. “We discovered that Mirai uses comparisons between the left and right sides, which is how we were able to design a substantially simpler network that also performs comparisons between the sides.”
These researchers found that the new model had performed with nearly the same success as the previous model in predictions up to five years in the future.
“We can, with surprisingly high accuracy, predict whether a woman will develop cancer in the next 1 to 5 years based solely on localized differences between her left and right breast tissue,” said Donnelly, per the news release. “This could have public impact because it could, in the not-too-distant future, affect how often women receive mammograms.”
Since the reasoning behind AsymMirai’s predictions is easy to understand, Donnelly believes the new model may become a valuable addition to human radiologists in making breast cancer diagnoses and risk predictions.
Breast cancer is the second most deadly type of cancer among women, following right behind lung cancer. A report from Texas Oncology predicted that 20,510 new cases of female and male breast cancer would be reported in the state last year, 3,503 of which would be fatal, as previously reported by The Dallas Express.
