Maximizing your “Return on Innovation” for AI

Adopting and implementing new technologies often leads one to question - “What is the return on investment?” When it comes to truly innovative solutions, that is a question not so easily answered.  Taking a look forward at the use of artificial intelligence in medical imaging and other areas of healthcare, one can see it is ripe to deliver on the hype of improved outcomes, workflow automation, and systematic efficiency gains. Though most of today’s AI solutions can’t deliver a nicely packaged return on investment calculation, this is not necessary when you look at the importance of your “Return on Innovation.” Most of the novel advancements in healthcare started without reimbursement and a clear understanding of who was going to pay for it.  The path forward clearly needs the boost that can be delivered when looking at the projections of physician deficits, growing global demand for better healthcare, and the aging population in advanced countries  An interesting place to compare how this works is what we are witnessing with 3D printing for medicine. Clear and tangible benefits have been documented and demonstrated by innovators, with savings in surgical cost associated with planning and materials, shorter surgical times, and better patient outcomes. The use of 3D printing is finding its way into many new applications and adoption continues to grow, yet, there is no CPT code for 3D printing or other means of reimbursement.  It has taken research money and philanthropy to get where we are. Only now it is starting to reach mainstream after the case has been made. This is what is meant by a “Return on Innovation”. The use of 3D anatomic models is now expanding and transitioning into Augmented Reality and Virtual Reality for certain use cases. Still yet, no traditional ROI can be used to justify the cost.  It’s all about understanding the intrinsic value of innovation and the need to apply both tangible and intangible benefits in your final equation.

AI and ML promise to deliver on an unimaginable scale versus something like 3D printing and AR.  However, this will only happen if it’s well-executed and with the intent of delivering better healthcare at a lower cost.  The health systems and practice groups who choose to adopt AI at this innovation stage will be the long-term winners and raise the bar in the highly competitive service sectors of medicine.  Innovators will quickly learn the benefits and be able to measure the right things and for the right reasons.  Quality must and will be a key driver in the process.  Those who don’t embrace AI and join the first movers will likely be left behind as they won’t be able to compete with more efficient and higher quality delivery systems. There will be a learning curve, and likely some setbacks, but as seen all too often, those who master innovation first will be able to set the pace while everyone else plays catch up.

There is still room for caution - the Hype Cycle is in full effect - so it’s important to remain pragmatic. After various debates and accepting the current state of AI and ML, it is well established that AI will not entirely replace the Radiologist in our generation. There are many facets to this discussion.  First is that AI needs to be validated and proven. The “ground truth” process used previously for Mammography CAD and similar applications was part of its eventual downfall. In practice, most radiologists ignored the markings due to high false positives and their lack of trust in the algorithms.  AI solutions that allow the Radiologist to engage with the results and participate in the learning portion of the engine will help create the needed trust and the necessary “belief system” for wider adoption and streamlined diagnosis.  This means a second change is needed. We cannot use the PACS and reporting workflows that were mostly designed around replacing film and paper as our pathway to AI adoption. We need to see innovation in the image viewers and tools available to interact with AI. Ideally, this effort will be aligned with the “Cockpit of the Future” initiative being sponsored by  the  National Institute of Standards and Technology (NIST) and The Academy for Radiology & Biomedical Imaging Research.

We have been hearing about “Big Data” for far too long, and to be honest, big data is not the answer; mostly because it was never the question. AI and ML need to be applied to problem statements, not ideals.  Some have already made huge investments and gotten no return, proving that hitching your AI strategy to an individual company with either no focus or too much focus on a single problem isn’t the answer either.  With the rapid advancements in machine learning and computing power, the true innovators know they need to try many AI engines and workflows across many departments and applied to many problems statements. They will need to leverage an AI platform that lets them experiment with their own tools and engines while comparing and contrasting the available commercial solutions without committing to a long-term purchase and support model used for capital equipment or departmental information solutions.  Cloud usage and crowd sourcing will be required to get the most out of these technologies. The health systems who innovate early will also need to collaborate in order to ensure their own leadership position can be maintained. Now is the time to make your initial investment, as those who do will see the greatest value and return on innovation in their organizations.

The 2017 RSNA Machine Learning showcase was just the beginning.  Look for the companies and researchers who participated there to find a way to collaborate and deliver a broad range of solutions that are integrated and adaptable.  2018 will be the year of real AI adoption in medical imaging, the proverbial ‘cat is out of the bag’ now and it’s to be expected that the investments and improvements in ML and AI will overshadow any of the minor gains we witnessed with the billions of dollars spent on meaningful use initiatives.