Researchers created a privacy-preserving method to adapt adult-trained radiology AI models for use in pediatric populations.
Key Details
- 1Only about 17% of FDA-approved AI radiology devices are labeled for pediatric use.
- 2Children represent less than 1% of public medical imaging datasets.
- 3The proposed approach allows AI models to learn from pediatric data at multiple hospitals without sharing sensitive patient information externally.
- 4Nearly 3,000 chest radiographs from 30 NICUs across 20 countries were used to train and test the model.
- 5The adapted model outperformed adult and single-institution pediatric models in identifying endotracheal tube tip location, with an average error of 0.67 cm (vs. 1.29 cm and 4.3 cm, respectively).
Why It Matters
This approach addresses the scarcity of pediatric imaging data and privacy constraints, enabling more equitable AI development. It also offers a scalable method to adapt models for other underrepresented patient populations in radiology.
Source
Radiology Business
Related News
•Radiology Business
AI Matches Radiologists in Prostate MRI Cancer Detection in Global Study
A global study finds AI performs comparably to radiologists in detecting prostate cancer on MRI.
•Radiology Business
RadNet AI Tool Reduces Unnecessary Prostate Biopsies by 19%
RadNet's DeepHealth AI tool can help avoid nearly 1 in 5 prostate biopsies while maintaining high cancer detection sensitivity.
•Radiology Business
Multi-Vendor AI Adoption Delivers Efficiency Gains in Private Radiology
A large-scale European study shows that adopting multiple AI tools in private radiology practices leads to significant efficiency improvements.