University of California, Berkeley Berkeley, California
Assistant Professor, Computational Precision Health Joint Assistant Professor, University of California, San Francisco
Developing targeted screening strategies to improve breast cancer risk prediction.
Risk prediction models can inform the use of targeted screening strategies to achieve earlier detection of breast cancer. To improve existing risk prediction models, Drs. Yala, Barzilay and their colleagues have developed a mammography-based deep learning model called MIRAI. This model was designed to predict an individual’s risk of developing breast cancer by analyzing multiple mammography timepoints and leveraging potentially missing risk factor information. In addition, MIRAI has the potential to produce breast cancer risk predictions that are consistent across mammography machines. Developing MIRAI involved utilizing a large dataset of mammographic information from Massachusetts General Hospital (MGH). It was tested across diverse patient data accumulated from seven hospitals in the United States and other countries. In this retrospective testing, the model was able to reliably identify high-risk patients, outperforming existing methods by a large margin. The goal of the current study is to test MIRAI in a prospective study, where patients predicted to be high-risk by this model are followed by MRI screening to assess its accuracy.
Drs. Yala and Barzilay’s research thus far suggests that MIRAI-based screening can offer broad, cost-effective, individualized, and equitable improvements in breast cancer risk prediction compared to costlier methods such as MRI. Their current work will be conducted at UMass Memorial Hospital (UMH) which serves a racially and socio-economically diverse group of patients—this will enable testing of MIRAI across these different subgroups. And it will be the first large prospective trial of MIRAI-based risk assessment technology in a diverse clinical setting.
The team will continue to recruit participants and anticipates a total patient accrual of 20,000. Ultimately, the results of the study could lead to redefining breast cancer screening guidelines and yield novel clinical protocols for managing high-risk patients.
Adam Yala, PhD is an assistant professor of Computational Precision Health and Electrical Engineering and Computer Sciences at UC Berkeley and UCSF. His research focuses on developing machine learning methods for precision health and translating them to clinical care. His previous research has contributed to three areas: 1) predicting future cancer risk, 2) designing personalized screening policies, and 3) learning encoding schemes for private data sharing. Dr. Yala’s tools have been deployed at multiple health systems around the world and his research has been featured in the Washington Post, New York Times, Boston Globe, and Wired. He obtained his PhD in Computer Science from MIT, and subsequently became a member of MIT’s Jameel Clinic and Computer Science & Artificial Intelligence Laboratory.
2022
The AI Screening Project supported by Zeta Tau Alpha Foundation
Massachusetts Institute of Technology Cambridge, Massachusetts
Please remember BCRF in your will planning. Learn More
