NVIDIA brings AI to ORs
NVIDIA launched open AI computing platforms for hospital operating rooms to automate logistics and real‑time monitoring — part of a broader move toward integrated 'AI factories' across healthcare. The same push toward real‑time, GPU‑driven workflows hints at downstream effects for specimen logistics, digital pathology adoption, and cross‑discipline data flow in labs. (pymnts.com) (genengnews.com)
NVIDIA’s Open‑H surgical video corpus totals 776 hours of footage collected from 35 organizations, forming the backbone dataset for the new healthcare robotics platform. (pymnts.com) Johnson & Johnson MedTech, CMR Surgical, PeritasAI and Proximie are listed as early adopters building on NVIDIA’s open OR stack. (pymnts.com) NVIDIA published Cosmos‑H as an open model family to generate physics‑aware synthetic surgical data for edge cases and rare scenarios, reducing reliance on proprietary video sources. (completeaitraining.com) NVIDIA’s DSX Air simulation service lets organizations create high‑fidelity digital twins of compute, networking and storage to validate AI factory deployments, shortening integration time from months to days or hours. (blogs.nvidia.com) An XRlabs first‑in‑human neurosurgical deployment used an exoscope feed routed through NVIDIA Holoscan with a Jetson AGX Thor and a custom SMPTE ST 2110 operator to run a multi‑hour clinical procedure with GPU‑resident, ultra‑low‑latency processing. (nvidia-holoscan.github.io) Oath Surgical’s January 27 partnership with NVIDIA ties OathOS to NVIDIA’s real‑time video analytics; Oath currently operates three surgical centers in Portland, is affiliated with about 20 ambulatory surgery centers and a network of roughly 175 surgeons, and has raised about $35 million to date. (healthcare-brew.com) NVIDIA and ecosystem partners position PeritasAI to train systems for instrument handling, sterile‑field coordination and broader OR logistics—use cases the company and coverage identify as lower‑risk, high‑volume hospital coordination tasks. (completeaitraining.com) Mayo Clinic’s Digital Pathology initiative, built with NVIDIA and partners, already aggregates roughly 20 million whole‑slide images linked to about 10 million patient records, and early foundation models used 1.2 million de‑identified slides from Mayo and Charité in published development work. (newsnetwork.mayoclinic.org)
