Nvidia Corp. reckons artificial intelligence has evolved to the point where it can be reliably used to help diagnose diseases and discover new drugs.

To enable these new capabilities, the maker of graphics processing unit chips and computers, today announced it’s collaborating with two groups, including the American College of Radiology on an open AI architecture and reference implementation, and the Accelerating Therapeutics for Opportunities in Medicine consortium on a new supercomputer platform.

The open AI architecture and reference implementation is aimed at radiologists, who specialize in using medical imaging to diagnose and treat diseases. The reference implementation makes use of the Nvidia Clara AI toolkit, which provides three key AI capabilities including AI-assisted annotation, transfer learning and federated learning. Radiologists can create new algorithms, validate them, share them with colleagues and adapt them for a range of diagnosis purposes, while also ensuring the data used to train them remains secure.

The Nvidia Clara AI toolkit is packaged in the new ACR AI-LAB software platform that’s freely available for radiologists to use and develop AI algorithms based on their own patient data. “This marks an initial stage of an extraordinary ACR Data Science Institute project that gives radiologists in any practice environment an opportunity to become involved in AI development at their own institutions, using their own patient data to meet their own clinical needs,” said Bibb Allen Jr., chief medical officer of the Data Science Institute at the American College of Radiology.

One of the key differences between this implementation and previous programs is that the AI models are brought to the patient data, rather than the data being moved to the model. The benefit is that any data used remains confidential, as it should. It also helps to improve the diversity of AI training, improves validation of the algorithms and helps to teach radiologists how to adapt AI models for their specific clinical needs, Nvidia said.

In a pilot project with the Ohio State University, radiologists there imported an AI model that was originally created by doctors at the Massachusetts General and Brigham and Women’s Hospital’s’ Center for Clinical Data Science. The OSU doctors then added their own data to the model in order to improve the original algorithm. In a matter of days, the doctors came up with an extremely accurate “cardiac computed tomography angiography model” that could be used for diagnosing heart disease.

Just as important, the model was created by radiologists that don’t have any programming experience, and without any confidential data being shared.

“Radiologists want to get involved,” Kimberly Powell, vice president of healthcare at Nvidia Corp., said in a press briefing. “They want to build algorithms that meet their needs.”

Moreover, said Richard White, chair of the department of radiology and Medical Imaging Informatics at the Ohio State University Wexner Medical Center, “enabling a network of artificial intelligence between hospitals will create more robust algorithms, greater efficiencies and likely lead to better patient outcomes.”

Nvidia also wants to help achieve better patient outcomes by means of new drug discovery. To that end its working with the ATOM consortium on a new AI-based platform that’s designed to accelerate the process.

The ATOM consortium was formed in 2017 in order to speed up the long, drawn-out process of creating new drugs. It said the platform crunches a variety of complex data, including physicochemical properties, in vitro assay results and anonymized human clinical data. With this, doctors can employ data-driven models and generative molecular design to aid in the design of new medicines, the effects of which can then be simulated within a computer environment.

ATOM’s platform runs on a supercomputer powered by Nvidia’s GPUs, and the organization said it will make this available to the medical research community to help it come up with new drugs faster. “Collaborating with Nvidia, we intend to advance the role of computation in drug design, reduce experimental bottlenecks, and speed up drug discovery,” said Barry Selick, governing board member of ATOM and vice chancellor of business development, innovation and partnerships at UCSF.

Analyst Holger Mueller of Constellation Research Inc. told SiliconANGLE that Nvidia is trying to push AI into the next phase of its life, which is specialization by industry.

“It’s good to see Nvidia making it easier to build AI apps, especially in healthcare where due to subject matter expertise and patient privacy it’s important that medical practitioners themselves do the building,” Mueller said. “Likewise partnering with forward-thinking institutions like ATOM further helps in validation and propagation of new best practices.”

The key question, he added, will be whether particular industries will need their own specialized hardware.

Photo: Jarmoluk/Pixabay