The history and future of healthcare at Nvidia, according to its healthcare VP

Each week, we schedule our rounds with Healthcare Brew readers. Want to be featured in an upcoming edition? Click here to introduce yourself.

About 25 years ago, when Kimberly Powell was studying electrical engineering at Northeastern University, she took an internship with a medical imaging company that was working to build more technology into radiology.

She worked closely with radiologists and saw their work change when updated imaging equipment provided closer looks into patients’ medical issues. The problem was the company was building its own graphics cards, a computer system that processes image data (and is the major backbone of AI).

So she called up tech company Nvidia to ask to use its technology because around that time Nvidia was moving from computer graphics into accelerated computing, which requires specialized software to compute.

“They said, ‘We have early evidence that healthcare is going to be a great user of this technology, but we don’t know how,’” Powell said. Then they asked her: “Would you like to join the company and help us figure that out?”

Powell, now VP of healthcare with Nvidia, sat down with Healthcare Brew at the JPM Healthcare Conference in San Francisco on Jan. 13 to share what the company has accomplished in the healthcare industry over the last 17 years and where she thinks innovation is moving.

This interview has been lightly edited for length and clarity.

What were Nvidia’s biggest accomplishments in healthcare over the last 17 years?

That first phase of accelerated computing was very much in two main areas. One was sensor processing, where every diagnostic imaging device, these are essentially sensors, and all that sensor data has to be processed and turned into what they call an anatomical image, [which is] an image that a human can read. Because the sensor technology was continuing to get better, they needed a different technology than the [central processing unit]. They needed accelerated computing because images need to be ready in real time. That’s where we had a lot of energy in those first five years at Nvidia.

Another major area was molecular dynamic simulation. If you go back into the 2010 time frame, one of [all super computers’] major workloads was molecular simulations. They’re trying to figure out, here’s a protein that’s causing a disease, and I want to find a molecule that fits in…That still persists as one of the most important computational steps in drug discovery because that physics simulation really allows you to understand how effective these drugs are going to be.

Over the last five and 10 years, we recognized that in order to push this industry even further, we needed to go beyond just accelerating the algorithm itself. In some cases we designed full computer systems. In the area of imaging and medical devices, for example, we architected a whole system and system software called Holoscan and IGX.

What’s coming in the next 10–15 years?

We describe the next wave of AI as physical AI, where AI is embodied in the physical world. We did a lot of work with GE Healthcare over the last 12 months around how we could make imaging more autonomous and how we could make every device not only be smarter but actually potentially act.

Imagine if any human on Earth could go get a lung screening scan just by walking into a room and being completely autonomously led through that experience. You could imagine that humans, instead of doing the task of just reading the image, are more thinking about patient needs and overseeing the process and really helping patients make decisions.