Cataract Surgery, at the Speed of Sight Real-time surgical AI, engineered onto the OR floor and not behind it.
A pilot demo is easy. A pipeline the surgeon can rely on, on live 4K video, is engineering.
- Sub-frame 4K inference achieved, with capture, pre-processing, inference, post-processing and overlay moved onto GPU-accelerated CUDA and TensorRT paths
- OR-local AI assistance enabled on the NVIDIA IGX Orin edge platform, removing dependence on external servers, network roundtrips and cloud availability mid procedure
- Patient video kept inside the operating room, processed on-device by design to reduce exposure of sensitive procedure video
- Overlay latency reduced by folding Non-Maximum Suppression into the inference flow instead of running it as a separate step
- Single edge-box deployment created, consolidating video input, inference and overlay output in place of a workstation and server stack
- Reusable surgical AI fabric established, ready to extend to phaco-tip tracking, IOL alignment guidance and complication prediction without re-plumbing the integration
Built on NVIDIA Holoscan's graph-based operator model, so the pipeline stays adaptable as cataract AI models evolve.
Download the case study to see how LTTS engineered a real-time surgical AI pipeline that runs inside the OR, not behind it.
Download Case Study