NSF Research Shows How AI can Reduce False Positives in Breast Cancer Screening
October is breast cancer awareness month and the National Science Foundation (NSF) is showcasing how federally funded artificial intelligence (AI) research has made advances in eliminating false positives in ultrasound breast cancer screening. While ultrasound technology is capable of detecting signs of breast cancer at a minimal cost and without exposing the patient to radiation, it isn’t always utilized as it has a much higher rate of false positives than other detection methods. But now, with NSF researchers using available ultrasound data to train AI systems, non-invasive ultrasounds may be used to more-accurately detect breast cancer.
Using the developing AI model, researchers predict ultrasound breast cancer diagnosis accuracy will increase from 92% to 96%, which will reduce the need for unnecessary, invasive biopsies. The AI tool will be further refined to include genetic factors to continue to increase the accuracy of AI-aided ultrasound diagnostics.
The research has been published in Nature Communications, providing a complete review of the summary. There, the researchers summarize, “Deep learning models trained on those datasets might not sufficiently learn the diverse characteristics of US images observed in clinical practice. This is especially important for US imaging as lesion appearance can vary substantially depending on the imaging technique and the manufacturer of the US unit system.”
As AI continues to advance and find new applications, engineers must consider how it can and will be used to improve existing technology, while also considering potential consequences. This study did consider how AI learning through datasets, as done with this research, could be a source of bias. It is important that considerations of bias be thought through at the outset of research as AI continues to find new applications.