Amy Kaplan

How open source AI helps doctors plan surgery on a child’s heart 🫀 At @ChildrensPhila (CHOP), Dr. Matthew Jolley’s team models a child’s heart in 3D before surgery: in one case, two repairs had failed until the model clarified the anatomy. 💡 TLDR: open-source AI turns a child’s heart scan into a 3D model in seconds instead of hours, and because the tools are open, other children’s hospitals can use them and build on them. How a scan becomes a 3D heart 👇 ➡️ The team trains AI on scans where researchers already marked the anatomy ➡️ On a new scan it predicts which regions belong to which heart structure (that’s segmentation) ➡️ It runs on MONAI, an open-source medical imaging framework NVIDIA co-founded ➡️ Doctors explore that anatomy in 3D and plan around one specific child’s heart A modeling step that took a skilled researcher about four hours now finishes in seconds. Why open source? Kids’ heart defects are too rare and too varied for traditional device companies to invest in. No single company has built these tools, and no single hospital could build them alone. So Dr. Jolley’s lab built them with the open-source community and made them free to use and build on. Stanford and Boston Children’s now contribute tools alongside CHOP, and more than 20 children’s hospitals in the US run cardiac modeling programs. Each team builds on the others’ progress. Next: CHOP is working with NVIDIA on physics simulations of how a device will deploy inside a specific child’s heart, with the long-term goal of bringing real-time simulations into clinical workflows. 📚Full story on the NVIDIA blog: blogs.nvidia.com/blog/childr…