Nvidia has introduced an open-source framework designed to train healthcare robots using physical AI, aiming to accelerate the development of surgical and diagnostic systems.
Key facts
- •Nvidia’s framework uses parallel simulation to cut training time, with one benchmark showing a reduction from over five hours to under two minutes using 8,192 environments.
- •CMR Surgical contributed approximately 500 hours of anonymized clinical data from its Versius Surgical Robotic System to the Open-H Embodiment dataset.
- •The framework combines classical physics modeling with generative AI to simulate complex scenarios like guidewires catching on vessel walls.
- •Johnson & Johnson MedTech is using the platform to build a digital twin of its endoluminal MONARCH system for kidney-stone procedures.
- •The framework is designed to help developers create evidence trails for regulatory bodies like the FDA by allowing inspection of internal physics assumptions.
Nvidia has announced a new Medical Physics Simulation framework, an open-source addition to its Isaac for Healthcare platform. The technology treats healthcare robots as physical AI systems, allowing them to learn through simulated physical interactions rather than relying solely on text or image data. By generating synthetic scenarios, the framework aims to overcome the scarcity of clinical data and the slow pace of real-world training for surgical and diagnostic robotics.
By the numbers
How the Simulation Framework Works
The framework utilizes a dual-modelling approach to simulate internal body environments. Classical physics simulation manages mechanical rules, such as how a catheter bends or how resistance is applied by vessel walls. Meanwhile, a generative AI component called Cosmos-H Dreams handles visual scene dynamics. By running these simulations in parallel on GPUs using Nvidia’s Warp and Newton libraries, developers can train robots across thousands of environments simultaneously, significantly reducing the time required for model development.
Industry Adoption and Early Use Cases
Several companies are currently utilizing the framework for various applications. CMR Surgical has contributed nearly 500 hours of anonymized clinical data to the Open-H Embodiment dataset, covering procedures like hernia repairs and hysterectomies. Johnson & Johnson MedTech is developing a digital twin of its MONARCH platform for urology, while XCath is focusing on endovascular autonomy. Other firms, including Inner Logic and Medtronic Structural Heart, are using the tools for device validation and catheter navigation research.
Regulatory and Clinical Considerations
While the framework accelerates training, Nvidia and its partners emphasize that these systems are currently in the training and research phase, not in clinical use. A primary challenge for healthcare robotics remains the governance requirement; regulators like the FDA require evidence trails explaining how a system arrives at its behavior. Nvidia argues that an open-source framework allows developers to inspect physics assumptions and build the necessary evidence for regulatory submissions, though it does not replace the need for clinical validation.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by Artificial Intelligence News.
