Open-source tools enable children’s heart simulations in seconds, breaking the limits of congenital defect surgery

Approximately one percent of live births are born with congenital heart defects, which exhibit a wide anatomical variability that makes each case unique. Pediatric heart surgeons have historically faced a structural dilemma, requiring them to address complex, exceptional anatomy using off-the-shelf devices and supports designed according to predefined standard measurements. At the Children’s Hospital of Philadelphia, Dr. Matthew Julie, a cardiologist and researcher at the hospital, led an approach that relies on open-source artificial-intelligence tools to build three-dimensional anatomical models of children’s hearts within seconds, after verification of this data previously required about four hours of manual work by a specialized researcher.

The system relies on the open-source medical framework MONAI, which NVIDIA helped found, together with specialized open-source software packages such as SlicerHeart, an extension built on the 3D Slicer platform for processing and analyzing three-dimensional medical images. By employing tools like MONAI Label and the Auto3DSeg model, neural segmentation networks were trained on pairs of images and previous models to match the accuracy of human experts.Reducing model extraction time from four hours to a few seconds has moved this technology from academic research to a daily clinical care standard.Surgeons use CT, MRI, and three-dimensional ultrasound images to build digital models, compare the placement of replacement valves, and study cardiac wall stress before opening operating rooms or performing catheterization. This modeling has proven its practical value in correcting complex deformities, such as ventricular septal defects, and succeeded in guiding the surgery of a child who had previously undergone two failed attempts in which conventional methods could not pinpoint the defect accurately.

Medical ambition does not stop at visual perception, but goes beyond it to test device behavior inside living tissue in advance using physical simulation engines. The team at the IDEA lab of the Morgan Center for Research and Innovation tests the Newton engine, an open-source physics engine based on NVIDIA’s Warp library for running simulations on graphics processing units. This engine was originally designed for training intelligent robots, but its integration into medical analysis tools allows understanding of the biomechanical properties of tissues and the behavior of hole-closing devices and catheter valves. This acceleration reduces the simulation time for cardiac interventions from a full night or four hours to roughly real-time, enabling physicians to compare multiple treatment options and make therapeutic decisions on the same day. The development also extends to linking these models with digital-twin environments via the OpenUSD standard, to integrate virtual-reality technologies and vision-and-language models, thereby allowing surgeons to interrogate cardiac anatomy and interact with it proactively.

This shift gains a highly significant economic dimension, as about 2.4 million people in the United States live with congenital heart defects, a group that is anatomically heterogeneous and numerically small compared with adult diseases, which makes them unattractive to traditional commercial investments by medical-device companies that prefer large markets with standardized solutions. Juli explains that breaking these economic constraints was not possible except through open source, as leading pediatric hospitals such as Boston Children’s Hospital (a leading pediatric hospital), which supports modeling for more than half of heart surgeries, around 500 cases annually, alongside Stanford (a leading pediatric center) and the Children’s Hospital of Philadelphia, which expects to complete 200 cases this year, are establishing a national alliance of more than twenty hospitals that share software infrastructure without reliance on a monopolistic commercial firm.

At the level of hospitals and specialized cardiac centers in the Gulf, Egypt and Jordan, this shift carries direct operational and financial significance. The biggest challenges of pediatric cardiac surgery in our regional centers have revolved around the high cost of licenses for closed-source software systems and the monopoly of device manufacturers over surgical planning software. This model shows that the most efficient investment is no longer directing large budgets toward proprietary solutions that lack the flexibility required for complex dissection, but rather building clinical and engineering competencies that combine biomedical engineering specialties with open-source programming skills and locally accelerated graphics computing. Catheterization and cardiac surgery departments in the region can adopt frameworks such as MONAI and SlicerHeart to develop custom diagnostic solutions for their patients, relying on their own clinical data without waiting for initiatives from global companies that often overlook rare cases because of their low commercial return.