Privacy-preserving training could expand medical AI for children while keeping data local

Researchers at Harvard Medical School developed a privacy-preserving AI training method that adapts adult medical imaging models for pediatric use by training sequentially on data from multiple hospitals without sharing sensitive patient information. This method outperforms existing pediatric models and could expand AI applications in neonatal and pediatric care despite strict privacy regulations.

This Wire brief sits within Fusion42's coverage of Healthtech Infrastructure, and 2 sources have reported it between 13 Oct 2026 and 18 Oct 2026.