Detecting breast cancer with artificial intelligence

Breast cancer screening currently involves radiologists reading mammograms to identify abnormal tissue for further assessment. Randomised trials have shown the benefits of screening in reducing mortality; however, some cancers are still missed at screening and progress to become life-threatening. Artificial intelligence (AI) has the potential to further improve the accuracy of mammographic reading and increase detection of these clinically-important cancers. In this project, an AI algorithm that has been shown to perform well in research settings will be applied to a population screening cohort (110,000 mammogram images from the BreastScreen WA database), and its accuracy will compared with human readers. Those results will also inform an economic analysis to understand whether using AI will be more or less costly than current screening processes, and whether any extra costs are outweighed by benefits of the technology. Finally, given little is known about how the community perceives AI for breast cancer screening, perspectives and preferences of Australian women will be explored. The project will therefore help shed light on whether and how AI can be used in ways that increase cancer detection, contribute the ongoing sustainability of population screening programs, and are acceptable to the community.

Project lead: Dr Luke Marinovich