Graphics processing unit technology developer Nvidia will integrate its Clara artificial intelligence (AI) toolkit into the American College of Radiology (ACR) AI-LAB software platform under an agreement announced April 8.
The ACR's Data Science Institute on April 5 announced AI-LAB, designed to be a platform that will enable radiologists and imaging facilities to develop their own AI algorithms. The ACR hopes to "democratize" artificial intelligence by offering support for radiology professionals and institutions throughout the algorithm development life cycle.
Under the deal between the ACR and Nvidia, the ACR will integrate the Clara AI toolkit directly into AI-LAB. Nvidia launched the Clara AI toolkit at the 2019 GPU Technology Conference as a platform to facilitate the training and deployment of AI algorithms. Clara AI includes 13 classification and segmentation AI algorithms and software tools; it also supports transfer learning, enabling existing AI models to be adapted to fit local variables.
The relationship follows a successful three-month pilot program with researchers at Ohio State University (OSU), the Massachusetts General Hospital (MGH), and Brigham and Women's Hospital's Center for Clinical Data Science. Researchers were able to take a CT segmentation algorithm developed at MGH and share it over the AI-LAB network with an OSU team that then adapted the algorithm for the unique needs of its patient population, according to Kimberly Powell, vice president of healthcare at Nvidia.
The collaboration between ACR and Nvidia is important because it helps radiologists get up to speed in AI algorithm development without having to become immersed in the minutiae of writing computer code, according to Dr. Bibb Allen, chief medical officer of the ACR Data Science Institute.
Allen believes that the arrival of AI in radiology will be as significant as the adoption of advanced modalities like CT and MRI. Most radiologists want to become proficient in AI but need tools that will help them do so.
"We think that the next generation of radiologists needs to be as facile with AI as we have become with CT and MRI," Allen said. There is going to have to be a lot of hands-on learning. If the only solution we provide is Python for Dummies, people will never do it."
In conjunction with the ACR/Nvidia announcement, GE Healthcare announced that its Edison artificial intelligence platform will be integrated into AI-LAB. The integration will allow GE customers to more easily develop and deploy AI algorithms, such as its AIRx automated workflow tool for MRI brain scanning, GE said.
