NVIDIA and 75 Healthcare Partners Power Future of Radiology

At RSNA, NVIDIA launches new software, announces new partners to improve quality, access and cost of care.
by Abdul Hamid Halabi

Editor’s note: The name of the NVIDIA Transfer Learning Toolkit was changed to NVIDIA TAO Toolkit in August 2021. All references to the name have been updated in this blog.

Artificial intelligence research in radiology has shown great promise to improve quality, access and cost of care. However, it’s going to take an army of allies to bring this research to clinical practice. And that’s why NVIDIA has been hard at work expanding our ecosystem of healthcare partners.

We’re now working with 75 partners to apply AI to healthcare. It’s a number that grows every month. And it includes a diverse collection of medical centers, medical imaging companies, research institutes, healthcare startups and healthcare providers.

Many of these partners will be joining us at the Radiological Society of North America convention, taking place this week in Chicago. In addition to showcasing some of our work with them at RSNA, we’re announcing several important developments:

  • Availability of the NVIDIA Clara software development kit (SDK)
  • Unveiling of the TAO Toolkit for Medical Imaging and AI-Assisted Annotation SDK
  • Ohio State University is partnering with NVIDIA to build the first in-house AI marketplace using the NVIDIA Clara platform
  • National Institutes of Health is partnering with NVIDIA to bring AI tools to clinical trials

Intelligent Imaging: Clara SDK Available Now

With the newly available Clara SDK, developers can easily take advantage of any GPU platform they have to deploy AI, visualization or compute-intensive applications such as image reconstruction.

NVIDIA GPUs have played a key role in medical imaging for more than a dozen years. Diagnostic imaging modalities rely on our GPUs to deliver real-time, state-of-the-art image reconstruction. This includes iterative reconstruction to reduce radiation dosage of CT scans, compressed sensing to decrease scan time in MRIs, and software beamforming to increase image quality from ultrasounds.

Yet AI is capable of making image acquisition even better. Imaging instruments need AI to ensure the highest quality images are being acquired. Imaging companies like United Imaging, Fujifilm and Canon have all deployed NVIDIA DGX supercomputers as the AI infrastructure to accelerate their development of AI.

The Clara SDK is part of the open NVIDIA Clara platform, which enables the medical imaging industry to create and deploy advanced imaging applications and AI-enabled workflows.

MGH & BWH Center for Clinical Data Science has adopted the NVIDIA Clara SDK as a part of their AI deployment strategy. They have developed an Abdominal Aortic Aneurysm detection model and are deploying it on the Nuance AI Marketplace powered by NVIDIA Clara.

“If radiology is to benefit from the thousands of new AI applications being developed, we will need to have a clear path to deployment at a broad spectrum of clinical and imaging centers. This deployment path is key to a scalable adoption of AI in radiology,” said Mark Michalski, executive director at MGH & BWH Center for Clinical Data Science.

Learn more about the Clara SDK collection of GPU-accelerated software tools, libraries, AI engines, containers and sample applications.

Chart describing AI imaging chain

Radiology Workflows Need Thousands of Algorithms

Changing the practice of radiology is going to require thousands of applications. The demand for AI applications and the need to adapt them for each institution’s patients, machines and practice is why over 50 leading healthcare institutions — including MGH, BWH, NIH, UCSF, OSU, Mayo and KCL — have invested in NVIDIA DGX systems to develop AI applications.

To boost the radiology industry’s ability to build and adapt AI applications, NVIDIA has announced two key technologies:

  • The AI-Assisted Annotation SDK enables radiologists to unlock the value of their data 10x faster than traditional annotation methods.
  • The TAO Toolkit for Medical Imaging enables physicians to customize and adapt AI applications to their own patients. This is critical because every radiology practice is unique, with its own instruments, protocols and patient demographics.

“At OSU we understand the importance of these tools. Data curation is one of the major bottlenecks in the algorithm development lifecycle. This is particularly true in medical imaging due to the inherent complexity of the data and limited availability of highly trained annotators,” said Luciano Prevedello, chief of the Imaging Informatics Division at the Ohio State University Wexner Medical Center.

“Techniques such as transfer learning employed by the toolkit can significantly decrease the number of images required for training without detriment to algorithm performance,” Prevedello continued. “This, in combination with a more efficient data curation process that leverages AI to prepare the cases, will open the doors to a new era in algorithm development.”

Ohio State University Builds First In-House AI Marketplace

The Ohio State University Wexner Medical Center, a leading-edge academic medical center and university, is the first U.S. partner to adopt the NVIDIA Clara platform to build an in-house AI marketplace for clinical imaging.

OSU’s AI marketplace will allow radiologists to quickly apply deep learning and machine learning to their workflows.

“The rapid adoption of artificial intelligence has opened new opportunities in medical imaging,” said Dr. Richard White, chair of the department of radiology, Medical Imaging Informatics at the Ohio State University Wexner Medical Center. “Working with NVIDIA, we’ve streamlined the process of integrating AI into the workflow, which will lead to improved patient outcomes.”

OSU will deploy deep learning and machine learning to improve clinical responsiveness in urgent conditions, like detection of a brain hemorrhage or coronary artery disease. These algorithms could be integrated into many clinical workflows, such as an early warning system in an ER department, a worklist optimization in a radiology lab or as a diagnostic assistant in the reading room.

This points to another benefit: by standardizing on a deployment platform, organizations can potentially share and integrate all of the great AI applications being created by this exploding ecosystem.

National Institutes of Health Brings AI Tools to Clinical Trials

NVIDIA is also partnering with the National Institutes of Health, which runs the largest research hospital in the U.S., conducting over 1,600 trials a year.

NVIDIA will co-locate researchers and engineers with clinicians at the NIH Clinical Center. Our initial projects together will investigate AI tools to streamline clinical trials for brain and liver cancer.

The co-development will also focus on developing AI tools that combine imaging, genomic and clinical data to deliver precision medicine to cancer patients. This will be delivered through a specialized AI data-centric platform and deep learning-based radiomics.

“Applying a powerful tool such as deep learning to medicine will require a truly multidisciplinary team of physicians, hospitals and computer scientists to work together to help realize the potential of computer models for medical imaging, and to help develop predictive imaging biomarkers,” said Dr. Elizabeth Jones, director of the Radiology and Imaging Sciences Department at the NIH Clinical Center.

Currently, radiologists use manual measurements of tumors to determine cancer staging based on existing guidelines. AI will transform the process by automatically characterizing and measuring tumors in ways that might be imperceptible to the casual observer.

Further, AI has the potential to improve the accuracy of cancer staging by incorporating data beyond the size of the tumor and other currently used staging criteria. Novel imaging biomarkers, discovered by AI, can be used in clinical trials to get us a step closer towards predictive and personalized precision medicine.

To bring AI to radiology across the globe, we need to get radiologists involved in the creation and adaptation of the algorithms for their patients. It’s also important to give them standardized ways to share and integrate these breakthroughs with their colleagues and enable them to perform onsite data analysis with less regulatory or privacy risk.

Intelligent instruments and automated workflows are a reality. NVIDIA is partnering with industry thought leaders to enable radiology to cross the AI chasm through the NVIDIA Clara platform.