Skip to content Skip to navigation

LIVE from RSNA 2016: Rasu Shrestha, M.D. on Machine Learning and Other Paths to the Future

November 29, 2016
by Mark Hagland
| Reprints
UPMC’s Rasu Shrestha, M.D., on the technologies that will support the future of radiology

Rasu Shrestha, M.D., the chief innovation officer at the Pittsburgh-based UPMC health system, serves as the chair of the Informatics Scientific Program Committee at the Radiological Society of North America. In that role, Dr. Shrestha has led the discussions that have created the official theme each year for the past two years, for the imaging informatics content at the annual RSNA Conference. Last year, the theme was 3D printing; this year, it is machine learning.

Dr. Shrestha took out time on Nov. 29 during the frenzy of activity at RSNA 2016, being held at the McCormick Place Convention Center in Chicago, to speak with Healthcare Informatics Editor-in-Chief Mark Hagland, about the current state and future prospects of radiology practice and of imaging informatics. Below are excerpts from that interview.

With regard to this year’s theme, where are we right now in the industry with the concept of machine learning?

I think we’re in a pretty early but really promising and interesting, space. If you look at healthcare IT in general, we’re data-rich and information-poor. We have 9.7 petabytes of data, doubling every 18 months, at UPMC. So how do you “de-noise” the data? Data is an asset. But insights, information, knowledge, are what you really want. So moving from data to information, and onto knowledge and ultimately insights, and making those insights relevant and actionable at the point of decision-making, is really important. It could be decisionmaking on the part of a clinician, an administrator, a patient. How do we effect behavior change? At the end of the day, that’s what we’re trying to do.

Machine learning is important in this context: historically, we dealt with analog-based data. Now, machine learning gives the opportunity to look at data at scale: millions of similar images, with millions of similar types of cases looked up before, with best practices around certain types of pathologies. So how do we augment clinical decision-making? That’s really the promise of machine learning.


Rasu Shrestha, M.D.

What does that kind of leveraging of machine learning and data look like in radiology?

We’ve traditionally been very volume-based in radiology. How do we move from healthcare being interpretation-centric and volume-centric, to being value-centered and population-centric? We need to use the data in ways that help us to holistically addressing the problems around the individual patient. What we need to do is to be able to work with not just the presenting symptoms, or reasons for exam, such as “headache” or “cough”—often, that’s all the information you have. And even as radiologists evolve, and we need to become less diagnosticians and more consultants, we need to make sure that technology becomes an asset, not an impediment. In many ways, as we’ve been implementing EHRs [electronic health records] and PACS [picture archiving and communications systems] and CAD [computer-aided diagnosis], we’ve been replicating analog processes. But with the advent of machine learning, analytics, population health, and value-based healthcare in general, we’re able to say, it’s not just about doing digital, but being digital.

Doing digital was our goal in the last 20 years. Now that we’re all digital, how do we leverage the fact that we have all these assets? Moving from doing digital to being digital requires a mindshift. At UPMC, we’re developing algorithms to support clinicians, and are looking to streamline workflow around oncology and elsewhere. And we’re looking at population health, and wanting to use insights to help patients, including those who might come to us with different types of disease acuity. So that’s what we’re doing, is using technology to move us forward.

Are radiologists, other physicians, and healthcare IT professionals, coming to a level of awareness that will enable them to be able to use these tools?

They are coming to it, and the resistance to doing so is much lower. Primarily, that’s because the delivery of care is changing. And they know that it’s important to either change or be changed, right? That’s critical. For us at UPMC and the things we’re doing at UPMC Enterprises, we don’t just want to be a passenger in a seat, we want to be in the driver’s seat on the bus. We want to lead change. And I think we’re a couple of years ahead of the marketplace in general, because we’re an integrated provider-payer system. We’re not just talking about value-based care; we’re doing it already. And that’s the beauty of some of the things we’re innovating on. As we’re creating change, we’re implementing these things in a marketplace that’s hungry for solutions. And the concept of the living lab becomes very important. You can’t innovate in a sterile office, disconnected from care processes. So if you’re able to do it in a learning lab-type situation, the possibility of enhancing care becomes that much more likely.

Are vendors stepping up to the plate to deliver what we need, in terms of the tools that will be required to get us into the future of healthcare?

Pages

Get the latest information on Medical Imaging and attend other valuable sessions at this two-day Summit providing healthcare leaders with educational content, insightful debate and dialogue on the future of healthcare and technology.

Learn More

RELATED INSIGHTS
Topics