Artificial intelligence (AI) is still gaining momentum in 2019 as researchers build bigger and better algorithms and vendors release new, state-of-the-art solutions. For healthcare providers, however, the world of AI can still be intimidating in a lot of ways; if you haven’t already started exploring the potential of AI, where do you begin? What’s the first step?
Ran Rumianek, Change Healthcare’s executive director of AI and growth, has worked with countless providers and vendors over the years as AI technologies have continued to evolve. He spoke with us about what radiology departments should do when they begin developing an enterprise imaging strategy, and some of the challenges they may face along the way.
There is a lot of activity and hype surrounding the use of AI in healthcare, especially in medical imaging. Do you have any advice for healthcare providers who have yet to embrace AI technologies?
There is potential for AI to help improve clinical, operational, and financial outcomes in imaging and healthcare overall, though we are still in the early stages of that trend. There are a growing number of AI products available now, and it’s only going to increase in the years ahead as more and more algorithms are developed and gain regulatory approval.
Providers need to remember that while the data and AI revolution will not happen overnight in imaging and healthcare, it will likely have a growing and significant impact over time and they need to prepare for that as part of their enterprise imaging strategy. Most enterprise imaging customers we work with do see that as one of the key pillars in their strategy. This strategy starts with the data foundation of the organization—how to consolidate the data assets across the organization and have the enterprise imaging platform, tools, and environment that will allow enterprises to leverage their data.
What are some key challenges providers and technology vendors need to overcome to unlock the potential value of AI?
The key point to keep in mind is that providers and clinicians need solutions, not algorithms or technology. AI is a powerful technology that can help solve some problems much better, but for it to become useful, it needs to integrate into solutions that are easy to use and help providers do their job better. These solutions should be improving workflows, enabling better decisions, saving time, and limiting wasteful spending. And this part of AI workflow integration, which we refer to as the “last mile of AI” is sometimes harder than the “first mile” of collecting the data and developing the algorithms.
Currently, there are a few dozen AI algorithms and applications available from different vendors and organizations. In the next years, this number will grow significantly. Most PACS systems have not been designed for this type of ecosystem, and it will be a challenge to scale and to integrate these algorithms in a simple and effective manner. This requires a whole new approach—not just patching the current solutions. The next generation enterprise imaging solutions will be purposely designed for this era of data and AI, and will be able to effectively transform data to insights and to better workflow and decisions.
Another challenge is simplifying the business models around AI so providers can leverage the best capabilities available from different vendors in the right price point.
What is the role of the cloud when it comes to AI and enterprise imaging?
I believe that the transition to cloud native enterprise imaging solutions is a significant enabler to the evolution of AI in medical imaging, and the ability to leverage it effectively at scale. Consolidating your data to the cloud is one key aspect of strategic planning when it comes to preparing your organization to take advantage of AI and analytics. Once your data is consolidated into the cloud, as part of your enterprise imaging solution, you’ll have all of your data in a single environment where AI algorithms and other analytical tools can be leveraged as they evolve. In addition, you can easily grant access to data as needed instead of sending copies of the data from one system to another.
I find this transformation similar to the one we went through in the way we manage our personal photos. Back when we all used cameras with film, there were not so many photos and we could manage them quite easily with photo albums. Then, when the digital cameras appeared, we started managing them in our computers, manually organizing them in folders and working to back them up frequently. When smartphones appeared, the amount of data and the need to share it exploded, so much so that cloud services began offering both storage and disaster recovery solutions. But this was just the start, and now, in cloud services like Google Photos and others, we can use AI to truly organize and leverage our photos—for example, we can search pictures of ourselves on the beach in a blue shirt and get immediate results. You can also easily and securely share these photos with others.
This is all similar to what the medical imaging industry will likely go through in the coming years as we continue unlocking the value of data.
Does implementing cloud solutions put patient information, and patient safety, at a greater risk?
Security and quality should obviously be a significant area of focus with every healthcare technology solution, and should be taken very seriously. In essence, though, we believe that the cloud is a significant enabler of greater security. For example, in cloud native systems, it is easier to keep the system up to date with the latest and greatest security updates and fix any gaps in security quickly. In addition, large cloud providers have access to the most skillful security experts in the world, something that isn’t always true for smaller hospitals, or even some larger hospitals. Finally, as we discussed, the ability to grant access to certain images rather than sending actual copies leads to better overall security.