At the Pediatric Academic Societies 2025 conference, American researchers presented findings supporting the idea that artificial intelligence can significantly improve the accuracy of identifying cases of physical abuse of children in hospitals.
The AI model analyzed more than 3,000 cases of reports involving children under 10 years old—mostly those under two—in seven U.S. medical institutions from 2021 to 2022. Standard diagnostic codes were incorrect in 8.5% of cases, meaning that about 1 child in 10 remained undetected. The new system reduced this figure fivefold to 1.8%.
Unlike similar systems in Canada and the UK, the American model is designed for use in real practice rather than only for statistics.
When signs of violence may be subtle, algorithms help find repeated but disconnected signals. Timely identification of abuse can change a child’s entire life path—and now the chances of that may be greater.
Source: Neuroscience News
URL: https://neurosciencenews.com/ai-child-abuse-detection-28741/
Source date: 2025-04-26T13:38:46.413Z
