New AI model predicts cancer risk based on breast asymmetry
Researchers have developed a new artificial intelligence (AI) model that reviews mammogram images and accurately assess a woman’s five-year risk of developing breast cancer. The study outlining the model’s creation is published in Radiology. [1]
The new model, AsymMirai, is based on the popular Mirai deep-learning algorithm, developed by researchers at the Massachusetts Institute of Technology to predict future cancer. While the AI has a proven track record of accuracy, Mirai’s methodology is not well understood, which raises questions about exactly how much radiologists should rely on its assessments and diagnoses.
For this study, researchers aimed to develop a simpler model that is less complicated but also retains the accuracy of Mirai.
"Mirai is a black box—a very large and complex neural network, similar in construction to ChatGPT—and no one knew how it made its decisions," lead author of the study Jon Donnelly, a PhD student at Duke University said in a statement. "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."
The direct comparison study pitting AsymMirai against its forbearer was conducted using 210,067 mammograms from 81,824 patients (mean age 59.4 years), taken from January 2013 to December 2020. AsymMirai identified breast asymmetry as a clinical marker for cancer risk, and on that basis performed nearly as well as the more-sophisticated Mirai in assessing risk, with correlations of 0.6832 for 1-year risk prediction and 0.6988 for 4–5-year prediction.
"Previously, differences between the left and right breast tissue were used only to help detect cancer, not to predict it in advance," Donnelly said. "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.”
AsymMirai exhibited an area under the curve (AUC) of 0.79 for 1-year breast cancer risk prediction, slightly lower than Mirai's AUC of 0.84. Similarly, for 5-year risk prediction, AsymMirai achieved an AUC of 0.66, compared to Mirai's AUC of 0.71. Notably, in a subgroup analysis of 183 patients where AsymMirai consistently identified specific tissue features over time, it demonstrated a 3-year AUC of 0.92, demonstrating its potential for long-term risk assessment.
Further, Donnelly and his colleague wrote, these results underscore the clinical significance of breast asymmetry, suggesting bilateral dissimilarity is a valuable imaging marker for assessing breast cancer risk.
"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," Donnelly added. "This could have public impact because it could, in the not-too-distant future, affect how often women receive mammograms."
The full study is available at the link below.
Chad is an award-winning writer and editor with over 15 years of experience working in media. He has a decade-long professional background in healthcare, working as a writer and in public relations.