In a first-time collaboration between NVIDIA and spinout company Hypervision Surgical Ltd from the School of Biomedical Engineering & Imaging Sciences, pioneering computer vision technology has been used in an-vivo surgery study at IHU Strasbourg, France, to provide tissue-analytics information previously invisible to the human eye. It is the first time NVIDIA technology has been used in surgery.
Hypervision Surgical uses Artificial Intelligence (AI) computing and GPU-accelerated high-performance computing to extract real-time tissue analytics from complex hyperspectral imaging data to assist surgical decision-making. The NVIDIA Clara Holoscan software development kit, combined with the NVIDIA Clara AGX developer kit, provides the Hypervision team with edge computing resources and an AI-driven software development framework for building their technology platform.
The first example of Hypervision Surgical’s Hyperspectral Intelligence augmenting the doctor’s vision is a quantitative tissue perfusion “heatmap” that can be provided alongside conventional red/green/blue colour information, allowing for an intuitive display of tissue-analytics information for surgical decision-making. For instance, perfusion information of bowel anatomy can help surgeons understand optimal resection margins during colorectal (colon/rectum) surgery thus having the potential to prevent serious complications for patients.
As part of the in-vivo study, Hypervision Surgical’s technology was used to guide a “keyhole” surgical procedure and provide real-time conventional colour imaging and tissue perfusion heatmaps side by side for visible anatomies. Real-time, high-resolution hyperspectral imaging requires processing live data streams. Sampling tens of spectral bands across the visible spectrum, rather than the conventional three red/green/blue channels, results in a much more detailed view of the tissue’s spectral response. However, this rich data stream needs to be processed quickly to avoid any delay in delivering and displaying information during the critical surgeries.
Hyperspectral Intelligence is ideally configured for taking advantage of GPU parallelism which leverages dedicated hardware engines for recording and playback of data streams without impacting the performance of the real-time live view for surgery. To accelerate the commercial deployment of its solutions, Hypervision Surgical recently joined NVIDIA Inception, a free program that nurtures cutting-edge startups, which gives it prioritised access to computing technology and technical training.
As a spin-out from King’s College London, Hypervision Surgical’s collaboration with NVIDIA also bolsters the work between King’s College London and NVIDIA to build the gold standard for AI infrastructure in the clinic.