AI Algorithms Excel in Predicting Breast Cancer Risk
Artificial intelligence (AI) algorithms have demonstrated superior performance in predicting five-year breast cancer risk compared to traditional clinical risk models, offering potential benefits in personalized patient care and prediction efficiency.
AI vs. Clinical Risk Models
In a large-scale study, AI algorithms surpassed the standard clinical risk model for breast cancer risk prediction. Clinical risk models rely on various data sources, whereas AI can extract additional mammographic features, enhancing accuracy and predictive capability.
Study Design and Methodology
The study analyzed thousands of mammograms, focusing on negative screening 2D mammograms from Kaiser Permanente Northern California in 2016. A sub-cohort of 13,628 women and 4,584 patients diagnosed with cancer within five years were included, with follow-up until 2021.
AI Algorithms Outperform Clinical Risk Models
Using 2016 screening mammograms, five AI algorithms, including academic and commercially available ones, generated risk scores for breast cancer over a five-year period. All five algorithms outperformed the BCSC clinical risk model, indicating their ability to identify missed cancers and breast tissue features that predict future cancer development.
AI Predicts High-Risk Interval Cancers
Certain AI algorithms excelled in predicting high-risk interval cancers, which are often aggressive and require additional screening or follow-up imaging. AI predicted a higher percentage of cancers compared to the BCSC risk model, offering potential improvements in early detection and intervention.
Integration of AI in Breast Cancer Risk Assessment
AI-based risk models using mammogram data provide practical advantages over traditional clinical risk models. Integrating AI-generated future risk scores into radiology reports could enhance personalized patient care, enabling the provision of precise, individualized medicine on a national scale.