Reply to: Transparency and reproducibility in artificial intelligence

More generally, widely releasing data considerably alters the risk-benefit calculus for patients, so institutions must be thoughtful about how and when they do this. Because of these considerations, large medical image datasets with associated breast cancer outcomes are rarely made openly available3...

Full description

Saved in:
Bibliographic Details
Published inNature (London) Vol. 586; no. 7829; pp. E17 - E18
Main Authors McKinney, Scott Mayer, Karthikesalingam, Alan, Tse, Daniel, Kelly, Christopher J., Liu, Yun, Corrado, Greg S., Shetty, Shravya
Format Journal Article
LanguageEnglish
Published London Nature Publishing Group UK 15.10.2020
Nature Publishing Group
Subjects
692/700/1421
692/700/1421/1770
Algorithms
Artificial Intelligence
Breast cancer
Deep learning
Humanities and Social Sciences
Liability
Mammography
Matters Arising
Medical equipment
Medical imaging
Medical research
Medical screening
Medical technology
multidisciplinary
Privacy
Reproducibility
Reproducibility of Results
Science
Science (multidisciplinary)
Software
Online AccessGet full text

Cover

More Information
Summary:More generally, widely releasing data considerably alters the risk-benefit calculus for patients, so institutions must be thoughtful about how and when they do this. Because of these considerations, large medical image datasets with associated breast cancer outcomes are rarely made openly available3-5. Because liability issues surrounding artificial intelligence in healthcare remain unresolved8, providing unrestricted access to such technologies may place patients, providers, and developers at risk. [...]increasing evidence suggests that a model's learned parameters may inadvertently expose properties of its training set to attack; how to safeguard potentially susceptible models is the subject of active research9.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ObjectType-Correspondence-1
ObjectType-Commentary-2
content type line 23
ISSN:0028-0836
1476-4687
1476-4687
DOI:10.1038/s41586-020-2767-x