Researchers have published 1,000 X-rays of children in cardiac intensive care, annotated on a device-by-device basis, along with a tool that is already integrated in real time at Boston Children’s Hospital.
In paediatric and cardiac intensive care, children undergo numerous chest X-rays to monitor their medical devices, fluid build-up in the lungs and wall oedema. However, these images are cluttered with probes, drains and catheters that obscure useful signs, and the scarcity of paediatric datasets is holding back the development of AI tools designed for adults.
A new dataset and a tool now in use
The team has first published PediaDeviceSeg-1K: 1,000 frontal X-rays of patients from a paediatric cardiac intensive care unit, in which each common device has been precisely segmented according to clinical criteria. The authors present it as the first public resource of its kind.
Building on this, PCXR-Copilot generates contours of the lungs and medical devices, can produce a view in which the devices are removed, and provides maps and quantitative indices related to oedema. The system is already operating in real time within Boston Children’s Hospital’s TWIST platform, demonstrating its feasibility in clinical settings.
Read the full article on the original source
Go to the source →