Nvidia today announced the launch of Clara Guardian, a smart hospital edge AI system that uses sensors to limit the spread of infectious disease and deliver things like operating room analytics and workflow automation. Clara Guardian and partners like AnyVision and Care.ai are currently deployed in 50 hospitals in countries like China, France, Italy, and Israel, covering 10,000 hospital rooms today.
In line with coronavirus-related concerns, Clara Guardian checks people for elevated temperatures, uses real-time computer vision for social distancing surveillance, and assists with contactless patient monitoring. In the future, Clara Guardian will also help with things like surgery analytics, handling tasks such as keeping track of equipment during an operation.
Nvidia insists smart hospitals can ensure cleaner environments and reduce the risk of spreading infectious diseases, including COVID-19. “If we can employ technologies like video and voice and use artificial intelligence to have a real-time active engagement to create actions and efficiencies for nurses, that’s a huge gain for health care professionals,” Nvidia VP of health care Kimberly Powell said in a call with reporters.
Nvidia CEO Jensen Huang today announced a series of news items central to the future of AI development via a prerecorded GTC keynote address. High on that list is the release of A100, Nvidia’s most powerful GPU to date, and the Ampere GPU architecture. The A100 promises 20 times better performance on training and inference compared to its predecessor, the Tesla V100. The A100 GPU will be available in public cloud environments and for the acceleration of Apache Spark 3.0 and AI-powered Nvidia SDKs like Merlin and Jarvis.
Clara Guardian works with Nvidia’s suite of EGX edge AI chips — ranging from the Jetson chip embedded in devices to T4 edge inferencing servers. Today, Nvidia also released Jetson Xavier NX and the EGX A100, the first edge chip to utilize the Ampere GPU architecture. The EGX A100 will be available in late 2020.
Nvidia also said it’s building a DGX A100 supercomputer for Argonne National Laboratory, which is a member of the COVID-19 High Performance Computing consortium. The supercomputer will carry out simulations to understand genetics and how molecules interact with each other.
Also new today from Nvidia’s health care team: pretrained models for lung segmentation and COVID-19 detection from CT scan imagery. The models are trained on data from China, Japan, the U.S., and Italy. Staff from the National Institutes of Health labeled the data, Powell said.
Quite a few pretrained COVID-19 detection models are available today, but in many places — including the United States — using CT scans for COVID-19 diagnosis is not the standard of care. But Powell urged the continued development of AI for CT scan imagery as a way to evaluate the severity of illness in a patient.
“These models can be used as building blocks [for] future things, like predicting whether a patient is going to go into an ICU, that will accelerate future research,” she said.