//
sign in
Post
by @danabra.mov
PostEmbed
by @danabra.mov
Record
by @jimpick.com
Record
by @atsui.org
+ new component
Post
In a new preprint, the authors identify 124 peer-reviewed papers which used provenance-unknown Kaggle data to train AI for stroke and diabetes risk. Some sets show impossible patterns, yet models of the models are already hitting clinics. #MedSky #MedAI
1mo
The models are designed to predict someone’s risk of diabetes or stroke. A few might already have been used on patients.
www.nature.com
Dozens of AI disease-prediction models were trained on dubious data
Scott McGrath