Enhancing AI Diagnosis
This is The Discovery Files, from the U.S. National Science Foundation.
The wide range of applications for artificial intelligence alone are impressive. When you pair up humans and AI, it could mean earlier and more accurate cancer detection.
Funded in part by NSF, researchers at New York University have learned, when artificial intelligence and human intelligence work together, they can make more accurate assessments of medical imaging tests.
Using an AI tool known as “Deep Neural Networks,” which enables AI to essentially “learn” how to assess data, researchers aimed to improve radiologists’ interpretation of the data. This increased understanding can help radiologists make more precise, health-related assessments, and help doctors make more accurate diagnoses.
Researchers used mammograms of breast cancer to compare radiologists and AI assessments to understand how AI and humans look at the images differently, in finding regions of potential concern.
While AI assessments relied more on the high number of features present, and in a wider area, radiologists focused on specific features and a narrower area.
The more radiologists understand about how these deep neural networks function, the better they can understand their limits, and how to incorporate them into a more accurate assessment.
This research illustrates how humans and machines can work in tandem, to make the world a healthier place for everyone, and potentially save lives.
Discover how the U.S. National Science Foundation is advancing research at nsf.gov.
“The discovery files” covers projects funded by the government’s National Science Foundation. Federally sponsored research — brought to you, by you! Learn more at nsf.gov or on our podcast.