Specializing in medical health! NVIDIA launched a new computing platform with AI computing power of 619TOPS

Zhi DongXi (public number: zhidxcom)

Author | ZeR0

Edit | Desert Shadow

In the field of health care, AI is playing an increasingly important role.

With the help of AI technology, doctors can sequence human DNA in just a few hours and predict the three-dimensional structure of DNA based on amino acid sequences; researchers can use computers to generate new drug candidates and test the efficacy of new drugs on target diseases.

This is the "AI+ medical" related progress shared by NVIDIA founder and CEO Jen-Hsun Huang in his keynote speech during the NVIDIA GTC conference this week.

In his view, the revolution in digital biology is particularly remarkable, with AI accelerating DNA sequencing, protein structure prediction, novel drug synthesis, and virtual drug testing. With AI pharmaceutical startups receiving more than $40 billion in investment over the past few years, the conditions are ripe for a digital biology revolution.

"This will be NVIDIA AI's greatest mission to date." Huang Jenxun said.

From the underlying hardware system to the software platform, NVIDIA continues to expand its layout in the field of medical and health care, providing dual support for computing optimization in this field.

On this basis, on March 23, NVIDIA launched a new AI computing platform with a computing power of 254-619TOPS to meet the needs of the medical device industry for deploying AI applications.

First, accelerate the introduction of AI technology into medical health

In recent years, more and more medical device companies have begun to integrate AI and robotics into medical devices to provide impetus for doctor diagnosis and biomedical research.

To that end, Nvidia launched the Clara supercomputer platform designed for medical imaging four years ago, which optimizes image acquisition and reconstruction, object detection and segmentation, and visualization capabilities based on existing medical equipment, enabling greater fidelity in medical imaging.

The NVIDIA Clara Software Development Kit (SDK) makes it easy for developers to deploy AI, visualization, or compute-intensive applications related to medical imaging on their GPU platforms. Nvidia has also created a transfer learning toolkit for medical imaging so that physicians can tailor and adapt AI applications to patient conditions.

In response to the extremely critical security and privacy issues in the field of "AI +medical", NVIDIA has partnered with King's College London in the United Kingdom to develop AN OPEN SOURCE AI framework MONAI optimized for the healthcare field. Based on PyTorch, it optimizes AI development in medical imaging through industry-specific data processing, high-performance training workflows, and advanced reproducible reference implementations.

NVIDIA Clara Holoscan, NVIDIA's intelligent computing platform for the healthcare industry, is based on NVIDIA AGX Orin, which seamlessly connects medical devices with edge servers, helping developers build, deploy, and manage a variety of AI medical imaging applications in a hybrid computing environment, accelerating high-speed I/O, physical processing, image processing, data processing, rendering, etc., so as to achieve a more real-time medical device workflow.

The Clara Holoscan SDK is designed for high-performance streaming data applications to build next-generation software-defined devices. It brings together pre-trained models and an extensible microservices framework that allows applications to be managed and deployed in device and edge data centers, pioneering the industry's software-as-a-service business model.

Second, the launch of a new AI computing platform, the computing power of up to 619TOPS

From robotic surgery to new approaches to studying biology, surgeons and scientists need medical devices to evolve into continuous sensing systems that can better drive disease research and treatment.

On March 23, 2022, NVIDIA released Clara Holoscan MGX, a platform for the medical device industry to develop and deploy real-time AI applications at the edge, that meets the necessary regulatory standards and meets IEC-62304 medical-grade specifications.

It combines Jetson AGX Orin industrial-grade modules, RTX A6000 GPUs, and ConnectX-7 smart NICs into a scalable AI platform that delivers AI performance of up to 254-619 trillion operations per second.

For high-throughput devices, ConnectX-7 is capable of providing up to 200GbE of bandwidth and a GPUDirect RDMA path for GPU processing, helping to speed up processing.

The Clara Holoscan MGX platform also integrates with the latest embedded security modules, which contain controllers responsible for monitoring critical operations, providing remote software updates and system recovery, and hardware root of trust for state-of-the-art embedded security.

In addition, the platform offers hardware reference designs based on long-life NVIDIA components, 10 years of long-term software support, and IEC60601 certification reports from embedded computing partners.

Medical device manufacturers can embed Clara Holoscan MGX directly or connect it to existing medical device installation platforms, enabling developers to accelerate AI deployment and regulatory approvals.

The Holoscan development platform is currently open to early experience customers, with an official launch date of May and a medical-grade readiness completed in the first quarter of 2023.

Several large medical device manufacturers and dozens of robotic surgical and medical imaging startups have developed based on NVIDIA's Clara Holoscan platform.

Third, the first batch of start-ups use the UK's strongest supercomputing to promote medical and health research

Nvidia is also using supercomputers to power research in the field of healthcare. It built the UK's fastest supercomputer, Cambridge-1, and undertook NVIDIA UK's AI and healthcare collaboration in industry and academia.

Officially launched last July, Cambridge-1 is an NVIDIA DGX SuperPOD cluster powered by the NVIDIA DGX A100 system, BlueField-2 DPU and NVIDIA InfiniBand network, one of the world's top 50 fastest computers, and 100% renewable energy.

Four members of NVIDIA's Inception project were selected as the first startups to visit Cambridge-1: Alchemy Therapeutics, InstaDeep, Peptone, and Relation Therapeutics. They will use the hashrate of Cambridge-1 supercomputing to accelerate drug discovery, genome sequencing, and disease research.

Alchemab Therapeutics is developing antibody drug discovery engines against diseases such as cancer, InstaDeep is providing AI-driven decision-making systems for the development of next-generation vaccines and therapeutics, and Peptone is developing a physics engine that helps analyze intrinsically disordered proteins (IDP) or precise structures of proteins that lack a fixed 3D structure, Oppenheimer, Relation Therapeutics is using recommendation system technology to uncover causal relationships in disease and advance new drug development.

Backed by supercomputers, British pharmaceutical giants AstraZeneca and Nvidia have developed the latest iteration of the natural language processing model MegaMolBART. The model can use AI to generate new molecules. The Transformer chemistry model can train a chemical language model with more than 1 billion parameters using the NVIDIA NeMo Megatron framework.

Conclusion: AI is helping more and more medical imaging and new drug development and innovation

From medical imaging, genomics to new drug development, the coverage and impact of AI technology in the field of healthcare is gradually expanding, enabling more and more breakthrough research and cutting-edge innovations that advance surgery, diagnosis and drug development.

The pursuit of "precision" of the medical and health industry, is becoming one of the world's largest demand for computing power industry, from the current trend, the integration of AI, robot technology of medical equipment is likely to become the main force of future medical equipment. Image processing and AI-accelerated computing are precisely what NVIDIA's GPU-related hardware and software products are good at.

Over the years, NVIDIA has continued to increase its investment in the field of health care, GPUs have been applied to almost all medical imaging equipment, and more AI software tools and reference applications that are convenient for developers and doctors are constantly updated. We look forward to seeing that with the support of these hardware and software resources, researchers can tap more potential for the combination of AI and medical and health research to change the future of medical care."