Heart of the Matter: How a Major Children’s Hospital Uses Open Source NVIDIA AI for Cardiac Care

Children’s Hospital of Philadelphia uses open-source stacks — MONAI, 3D Slicer/SlicerHeart, and NVIDIA tools — to turn CT, MRI, and 3D ultrasound into anatomically precise heart models in seconds. A workflow that once took a skilled researcher about four hours now fits routine clinical use.

About 1% of live births involve a congenital heart defect, each unique. More than 20 U.S. children’s hospitals run related modeling programs; Boston Children’s uses modeling in more than half of cardiac surgeries (~500/year). CHOP expects about 200 modeled cases this year.

We’re watching open medical AI do something the market underserves: tools for rare, diverse pediatric anatomies.

We’re at our most pro-tech when the evidence looks like this. Dr. Matthew Jolley’s line — one-of-a-kind kid, off-the-shelf device — is the whole problem. Modeling before the cath lab or OR is how you stop guessing with a child’s heart.

Machine learning on MONAI Label and Auto3DSeg did not invent pediatric cardiology; it collapsed the bottleneck after the lab had image-model pairs. Early wins on complex ventricular septal defects — including a child with two failed repairs until the 3D model clarified the hole — show why CHOP calls this standard of care for some cases.

The next step is physics: Newton on NVIDIA Warp aiming to cut device-deployment simulation from hours or overnight runs toward near real time, with paths into Omniverse/OpenUSD and VR plus vision-language queries. Useful AI that keeps clinicians in the loop and publishes open tools is the trajectory we want. The downside if this scales without rigorous validation is over-trust in a pretty model. I’m watching multi-center outcome data and how fast Warp/Newton biomechanics leave research and enter same-day decisions.

Roughly 2.4 million people in the U.S. live with congenital heart disease — historically too rare and diverse for traditional device-company tooling at the needed scale. CHOP’s IDEA Lab aims to extend the approach beyond cardiac care. Jolley joined CHOP in 2015 as 3D echo was arriving.