According to a recent study published in Radiology, Artificial Intelligence algorithms could be better at predicting the five-year breast cancer risk than the standard clinical risk model. In the latest study, thousands of mammograms were evaluated, and AI algorithms performed better than standard models.
AI is better than the standard breast cancer risk model
Researchers are exploring the use of artificial intelligence (AI) to enhance breast cancer risk assessment. Clinical models like the Breast Cancer Surveillance Consortium (BCSC) rely on self-reported information and other factors such as age, family history, childbirth history, and breast density to calculate a woman’s risk score.
Lead researcher Vignesh Arasu said that the availability and collection of data could limit the standard models. Therefore, to overcome these limitations, AI deep learning techniques can be used to extract a wealth of additional features from mammograms, providing a more comprehensive risk assessment.
In a retrospective study, Dr. Arasu analyzed data from negative 2D mammograms conducted at Kaiser Permanente Northern California in 2016. A total of 324,009 women were screened, and a random sub-cohort of 13,628 women was selected for analysis. The study also included 4,584 patients diagnosed with cancer within five years of the mammogram.
The study divided the five years into three segments: interval cancer risk (diagnosed between 0 and 1 year), future cancer risk (diagnosed between 1 and 5 years), and all cancer risk (diagnosed between 0 and 5 years).
AI predicted up to 28% of cancers
Several AI algorithms were used to generate risk scores for breast cancer based on screening mammograms in 2016. These algorithms included two academic and three commercially available ones, with all five outperforming the BCSC clinical risk score. Therefore, this suggests that AI is capable of identifying missed cancers and breast tissue features that contribute to future cancer development.
The AI algorithms showed particular proficiency in predicting patients at high risk of interval cancer, which is aggressive and may require additional screening or follow-up imaging. For example, when evaluating women with the highest 10% risk, AI predicted almost 28% of cancers, relative to 21% predicted by the BCSC.
