Health Tech, Hospitals, Artificial Intelligence

Nvidia says it has a solution for healthcare’s data problems

The chipmaker touted a new framework that would allow hospitals and pharmaceutical companies to collaborate on AI projects without sharing sensitive data. Nvidia said the framework is already gaining traction with hospitals and drug developers.


Healthcare leaders wouldn’t stop talking about one topic at the JP Morgan Healthcare Conference:  artificial intelligence. At Boston Consulting Group’s annual look-ahead for 2020, audience members rated digital, analytics and innovation as the top strategic priority for healthcare companies, even above pricing, access and M&A.

But healthcare companies face a unique challenge in that patient data is tied up in disparate health record systems, and subject to more stringent privacy laws. Chipmaker Nvidia — best known for its work powering the graphics seen in video games — touted a new solution on Thursday.

The company already has an existing line of technology products for imaging and genomics, called Nvidia Clara. In December, Nvidia unveiled a new application development framework that would allow researchers and drugmakers to collaborate without sharing patient data.

“The data remains in its residency. It never has to leave where it is. Whether that be a hospital, a pharma company, an insurance company,” Nvidia Vice President of Healthcare Kimberly Powell said, in her presentation at the conference that was livestreamed. “But you want to be able to collaborate.”

Called Clara FL (federated learning), the system would allow researchers and experts in their field to create AI applications. For example, algorithms might be trained on a pharmaceutical company’s data, but the data itself would not be shared with collaborators. Instead, the models from each dataset are shared and averaged together in one central model.

This process, of course, has some complications, mostly due to the distributed nature of the work.  But Powell said it had still won over users in healthcare.

“This is going to be truly transformational,” she said. “We already seeing massive adoption.”

For example, a consortium of pharmaceutical companies, including Johnson & Johnson and AstraZeneca, agreed to train their drug-discovery algorithms on each other’s data. The partnership, called the Melloddy project, gives them better odds at discovering new treatments while safeguarding their troves of data. A total of 17 tech and pharmaceutical companies are involved in the effort.

“In the pharmaceutical industry, their data is their IP. Yet they all know they need to innovate,” Powell said.

Academic medical centers have also taken up the technology, including UCLA Health, Massachusetts General Hospital and Brigham and Women’s Hospital’s Center, and four teaching hospitals in London.

Photo credit: Pixtum, Getty Images