Use of an AI Model to Predict 5-Year Breast Cancer Risk From Mammograms

By: Bryna Goeking
Posted: Monday, May 6, 2024

AsymMirai, a new radiology artificial intelligence (AI) model, demonstrated accuracy in predicting breast cancer diagnoses within 5 years from mammograms, according to a study published in the journal Radiology. The study, conducted by Cynthia Rudin, PhD, of Duke University, Durham, North Carolina, and colleagues, screened more than 210,000 mammograms from 81,824 individuals, yielding similar results to the state-of-the-art Mirai algorithm in predicting breast cancer.

“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 study coauthor Jon Donnelly, BS, a PhD student at Duke, in a press release from the Radiological Society of North America. “This could have public impact because it could, in the not-too-distant future, affect how often women receive mammograms.”

AsymMirai incorporates an interpretable module, with risk scores evaluated using the area under the receiver operating characteristic curve (AUC) and 95% confidence intervals. Focusing on local bilateral dissimilarity of breast tissue, AsymMirai achieved a 1-year AUC of 79%, a 3-year AUC of 68%, and a 5-year AUC of 66%. The model repeatedly highlighted the same issue in the mammograms of 183 patients, achieving a 3-year AUC of 92%.

Building upon the previous Mirai, an AI model whose inputs and operations are invisible, AsymMirai employs bilateral dissimilarity in a mammogram—a clinical tool previously used only in breast cancer detection. AsymMirai performed similarly to Mirai for 1- to 5-year breast cancer risk prediction, according to the researchers.

Disclosure: For full disclosures of the study authors, visit pubs.rsna.org.

Radiology

---