Know My Company
Tell us about your interaction with smart technologies like AI and cloud-based analytics platforms in the healthcare industry.
I think that’s the point: As a patient, I don’t experience the benefits of AI like I do when I use my cell, my smart home system or my self-driving car.
However, going back 10 years to 2009, one of the earliest users of neural networks on NVIDIA GPUs was a researcher at IDSIA in Switzerland, named Dan Ciresan, who applied deep learning to pathology images of breast cancer. Pathology images are trillions of pixels and just a couple hundred are maybe the cancer cells. Using deep learning by 2012, Dan and team won six international competitions in breast cancer detection. This was the kickoff to enabling the world of healthcare researchers and engineers to create AI algorithms applied to any health data.
How did you start in this space? What galvanized you to start at NVIDIA?
My first engineering career was for a startup in Boston that built the world’s first digital flat panel display system for mammography. I had the opportunity to work alongside researchers and radiologists at Mass General Hospital developing the first digital mammography systems. Seeing how technology could enable doctors to better serve patients was incredibly inspiring.
In 2008, with its invention of CUDA, NVIDIA was transforming beyond a world leader in computer graphics to an accelerated computing platform company. At the same time, R&D engineers developing next-generation CT, ultrasound, and mammography needed accelerated computing to bring the next wave of technology to the field of radiology. The timing was perfect.
I’ve been at NVIDIA for 11 years now and have the honor of leading the company’s healthcare business. We build hardware and software platforms leveraging accelerated computing, AI and visualization that power the ecosystem of medical imaging, life sciences, drug discovery, and healthcare analytics.
What is NVIDIA Clara AI and how does it benefit diagnostic radiology?
If you consider the entire practice of radiology, it’s super-complex, with instruments, workflows, systems, and operations. Radiology is a practice of deep specialization that covers tens of device types: CT, MR, ultrasound, tens of body systems, tens of organs within those systems, and many diseases that can affect any one of those organs.
Populations living longer and global access to healthcare are creating demand for radiology services that far outpaces the supply. It’s a global challenge. To meet that challenge, we will need thousands of AI algorithms to be developed and compute infrastructure to deploy AI applications.
That’s why we built the NVIDIA Clara AI toolkit, announced at our GPU Technology Conference in March. We need to enlist the world of radiologists to help build these thousands of AI algorithms, but they need tools to minimize the time needed to do so while maximizing the benefit to the radiologist’s workload. Clara AI provides an AI-assisted labeling capability that creates training datasets in one-third the time. It provides federated and transfers learning workflows so data can remain local and secure. And it provides a deployment workflow that makes it easy to validate and run AI algorithms in existing health IT infrastructure.
What is the state of AI for radiology technology in 2019? How much has it evolved since you first started here?
Over the last decade, radiology computing demands have seen a tenfold increase because of new detector technology, more sophisticated clinical applications using advanced image processing, and 3D. In the last three years, we have seen exponential growth and advancements in AI for radiology. For example, over 70% of medical imaging research is based on deep learning.
Because of the rapid change in computing approaches, radiology technology is quickly moving towards a software-defined environment where intelligent instruments and workflows will have the capability to continuously update and improve with the latest computational breakthroughs like accelerated computing and AI.
Tell us about your recent collaboration with the American College of Radiology and how it would transform the application of AI within radiology workflows?
On April 8, we announced a collaboration with the American College of Radiology to enable thousands of radiologists nationwide to create and use AI for diagnostic radiology in their own facilities, using their own data, to meet their own clinical needs. The Clara AI Toolkit is purpose-built to serve ACR members. We completed a successful three-month pilot program and now ACR is integrating the toolkit into its new ACR Data Science Institute ACR AI-LAB. The ACR AI-LAB is a free software platform that will be made available to more than 38,000 ACR members and other radiology professionals to build, share, locally adapt, and validate AI algorithms, while also ensure patient data stays protected at the local institution.
This will provide the technology and tools to connect a network of thousands of U.S. hospitals and every radiologist to use and deploy a pipeline that every hospital can implement into their existing infrastructure.
This collaboration will create a future of software-defined radiology — one where data and algorithms are continuously learning and improving everything from the sharing of models to the improvements of specialists using the system.
Where is AI-driven drug discovery heading? What would ATOM and NVIDIA deliver to innovative approaches in drug research and medicine improvements?
On average it takes six years to do drug discovery and enter into a clinical trial, not to mention hundreds of millions of dollars. Today the drug discovery process is time-consuming, sequential, and encounters high failure rates in clinical trials. The mission of the ATOM consortium, which is made up of industry leaders GSK, the U.S. Department of Energy’s Lawrence Livermore National Lab, the National Cancer Institute Fredericks National Lab and UCSF, is to use the massive amount of data that has been generated by the pharma industry over the last decade and create a parallel, in-silico framework to do drug discovery.
Also on April 8 (2019), we announced that ATOM and NVIDIA are combining forces to accelerate AI in drug discovery.
The three ingredients data, deep learning, and accelerated computing — are all available to the drug discovery industry and together with the ATOM consortium, we will help develop domain-specific tools and workflows that democratize the use of AI for all drug discovery researchers and accelerate the journey to precision medicine.
Tell us more about your vision into growing AI-driven healthcare and diagnostic technologies?
Modern AI and deep learning is the greatest technological breakthrough of our generation. The world of healthcare instruments, sensors, and health IT systems has created a data revolution, perfectly positioning healthcare to benefit from AI. Every major segment providers, payers, government, healthcare vendors can realize great innovation, efficiency, and advancement by becoming data-driven, software-defined and applying the state of the art in image recognition, speech recognition, natural language processing, and understanding to healthcare’s greatest challenges. Other industries like the consumer internet and self-driving cars have blazed a great trail toward these ends, but the last few miles are domain-specific. Our mission is to help the healthcare industry by creating domain-specific development and deployment platforms to innovation that improve the quality of care for every patient.
In the next year or so, you will see us contribute to AI in medical imaging, pathology, drug discovery, and genomics.
What is the biggest challenge to digital transformation in healthcare and diagnostics in 2019? How does NVIDIA contribute to a successful digital transformation?
Major challenges to digital transformation are making use of the data being generated and incorporating it as seamlessly as possible into the delivery of care.
NVIDIA helps solve the challenge of making data useful by building domain-specific tools for AI development. Clara AI is built for radiologists to get involved in the creation of AI applications. Likewise, it facilitates a hospital or a vendor to easily deploy this burgeoning growth of AI to any endpoint that makes sense: embedded in an instrument, in a hospital data center, or in the cloud. Our computing platform is the most ubiquitous, advanced, and future-proof computing infrastructure in the world. It will allow us to move to a software-defined industry and give every doctor and patient the benefit of AI.
What is the good, bad, and ugly about AI that you have heard or predict?
AI is not an application, it’s a technology, and I think this is misunderstood. If AI is a technology and the greatest of our time, then every company in healthcare should be building up their strategy to take advantage of it. It requires having a data strategy, an AI algorithm strategy, and a computing infrastructure strategy — three ingredients.
The second challenge not well internalized is AI is a computer writing software, and the more data you have, the more effective the software will become. This has two implications: first, you need a way to collect incorrect AI behavior in validation and deployment and feed that data back into the development pipeline, essentially a life-long learning loop. Second, because you are learning and improving all the time, your deployment infrastructure needs to be programmable and software-defined.
Thank you, Kimberly! That was fun and hope to see you back on AiThority soon.
Kimberly Powell is vice president of healthcare at NVIDIA. Kimberly Powell is responsible for the company’s worldwide healthcare business, including hardware and software platforms for accelerated computing, AI and visualization that power the ecosystem of medical imaging, life sciences, drug discovery and healthcare analytics. Previously, Powell led the company’s higher education and research business, along with strategic evangelism programs, NVIDIA AI Labs and the NVIDIA Inception program with over 4,000 AI startup members.
Powell joined NVIDIA in 2008 with responsibility for establishing NVIDIA GPUs as the accelerator platform for medical imaging instruments. She spent her early career in engineering and product management of diagnostic display systems at Planar Systems. Powell received a B.S. in electrical engineering with a concentration in computer engineering from Northeastern University.
NVIDIA’s invention of the GPU in 1999 sparked the growth of the PC gaming market, redefined modern computer graphics, and revolutionized parallel computing. More recently, GPU deep learning ignited modern AI — the next era of computing — with the GPU acting as the brain of computers, robots, and self-driving cars that can perceive and understand the world. Today, NVIDIA is increasingly known as “the AI computing company.”