How Children’s Hospital of Philadelphia Is Using NVIDIA’s Open Source AI to Personalize Heart Surgeries

At Children’s Hospital of Philadelphia (CHOP), researchers are working with NVIDIA and the open source community to use AI, 3D modeling and simulation to improve care for children with complex heart conditions. By turning medical scans into detailed models of a patient’s heart, the team can help doctors better understand anatomy and plan procedures before they begin. The work is also showing how open source technology can bring advanced tools into everyday clinical care.

According NVIDIA news, CHOP’s cardiac modeling service is built on MONAI — an open source medical imaging framework cofounded by NVIDIA

Congenital heart defects affect roughly 1% of live births, presenting unique anatomical challenges that historically forced surgeons to rely on ill-fitting, off-the-shelf devices. To solve this, Dr. Matthew Jolley and his team at Children’s Hospital of Philadelphia (CHOP) built SlicerHeart, an open-source extension of 3D Slicer. By training segmentation networks with NVIDIA MONAI Label and Auto3DSeg, the hospital transformed a tedious four-hour manual segmentation task into a process completed in seconds. This leap allowed high-precision 3D heart modeling to transition from an experimental research project into a routine clinical standard of care across complex pediatric surgeries.

Beyond static visualization, CHOP is advancing into real-time biomechanical simulation to predict how devices interact inside a patient’s body before any incision is made. Leveraging Newton—an open-source physics engine based on the NVIDIA Warp Python framework—researchers can run GPU-accelerated simulations of tissue stress and device deployment. This cuts simulation runtimes from up to four hours down to near real time, empowering clinicians to test multiple device configurations and make same-day treatment decisions for delicate hole-closure and valve-replacement procedures.

The innovation extends even further through digital twin technology built with OpenUSD and NVIDIA Omniverse. By coupling these framework simulations with virtual reality environments and embedded vision-language models (VLMs), surgical teams can intuitively examine and interact with a child’s virtual heart. Clinicians can query spatial relationships, test structural repairs, and evaluate hemodynamic blood-flow dynamics, gaining complete clarity on the patient’s unique anatomy long before entering the operating room.

Because the pediatric heart population is too small and varied to drive traditional commercial device investment, open source has become the essential workaround. Collaborating alongside institutions like Stanford and Boston Children’s Hospital, CHOP is helping lead a national consortium of children’s hospitals. By pairing broad clinical research with NVIDIA’s industrial-scale platform development, medical teams can bypass commercial limitations, democratize cutting-edge tools, and tailor life-saving interventions for one-of-a-kind kids.