CHOP Leverages NVIDIA's Open-Source AI for Pediatric Heart Modeling
AI-powered cardiac models reduce 4-hour workflows to seconds for congenital defect surgeries.
1Clinical-grade AI modeling
CHOP's service transforms existing CT/MRI/3D ultrasound scans into precise heart models in seconds using NVIDIA's MONAI framework. This replaces a manual 4-hour process per model, enabling routine clinical use. The system maintains quality equivalent to human experts while achieving 360x speedup. Over 20 US children's hospitals now use such modeling, with Boston Children's applying it to 500+ annual cardiac surgeries.
2Congenital defect challenges
1% of live births involve unique congenital heart defects requiring customized care. Traditional devices rarely match individual anatomies - a problem NVIDIA describes as 'one-of-a-kind kid vs off-the-shelf device'. Dr. Jolley's team at CHOP developed solutions specifically for pediatric cases, where small, complex anatomies lacked dedicated tools until their SlicerHeart extension for 3D Slicer software.
3MONAI ecosystem components
The open-source framework combines MONAI Label for active-learning annotation, MONAI Core for training libraries, and MONAI Deploy for clinical pipelines. Auto3DSeg implementation enables rapid segmentation. Maintained by 30+ institutions including NIH and Mayo Clinic under Apache 2.0 license, it uses PyTorch for medical imaging AI. This ecosystem allows continuous improvement through community contributions.
4Surgical impact cases
For ventricular septal defects, modeling became standard after proving decisive in complex cases. One child had two failed repair attempts before 3D visualization located the defect for successful surgery. CHOP now routinely models pre-operation, with 200+ cases expected in 2026. The technology has transitioned from research to standard of care for certain congenital conditions.
5Physics simulation integration
CHOP extends beyond visualization using Newton - an open-source physics engine on NVIDIA Warp. This enables GPU-accelerated device deployment simulations, reducing computation from hours to near real-time. Current work integrates biomechanics frameworks with SlicerHeart to predict device-tissue interactions. Closure device simulations are already implemented, with transcatheter valves next.
6Open-source economics
With 2.4M US congenital heart disease patients deemed too rare for commercial investment, open-source tools bypass traditional economics. SlicerHeart and MONAI allow multi-institutional collaboration without barriers. A national consortium is forming to build shared modeling infrastructure, with Stanford and Boston Children's contributing tools alongside CHOP.
7Future VR integration
CHOP is developing an Omniverse-powered coupler combining SlicerHeart with OpenUSD digital twins. This will embed vision-language models in VR environments, letting clinicians interact with cardiac anatomy pre-surgery. The long-term goal integrates real-time simulations into clinical workflows, potentially revolutionizing preoperative planning.
8Scalable precision medicine
NVIDIA's open-platform strategy (MONAI, Newton, OpenUSD) gives hospitals capabilities previously requiring corporate-scale engineering. By democratizing AI tools for heterogeneous populations, this approach addresses rare disease challenges through collaborative innovation rather than proprietary solutions - setting a template for other specialty medicine fields.
Original article: Unite.AI — CHOP Leverages NVIDIA's Open-Source AI for Pediatric Heart Modeling