In “Hit Refresh,” Satya Nadella, the current chief executive officer of Microsoft, Inc., gives us his fascinating take on the anticipated computing advances in store for this world . More importantly, he painstakingly details how Microsoft was losing the edge to the Googles, the Facebooks, and the Amazons of the computing world since it had conquered its original raison d’être —the founders’, Bill Gates’ and Paul Allen’s, goal of putting a computer on every desk in every home. Clearly, a revolution is taking place in computing, and the current intellectual gold rush for these technology companies lies in developing and harnessing artificial intelligence (AI) tools in various fields. Nadella states quite bluntly that Microsoft had to rediscover its soul—what made it unique. And Microsoft had to transform— hit refresh— in its “persistent quest for new energy, new ideas, relevance and renewal.”
Radiology is facing similar crossroads: we have conquered the digitization of medical imaging and picture archiving and communications systems (PACS). We have made what may have seemed impossible 30 years ago a reality. Many of us have not seen radiological images printed on film in over a decade, and some younger members of our profession may even wonder what that is. Today, studies with hundreds of images can be shipped across a hospital, a city, a country, or across the globe within seconds. If PACS was our end-goal, then we have arrived, and there are no more transformational challenges to tackle in radiology. However, Nadella’s insights on Microsoft are just as relevant for Radiology right now. PACS with its resultant abundant data streams has resulted in new challenges: without some help from technology, it will soon become almost impossible for humans to efficiently and effectively interpret all the images that our advanced acquisition devices can rapidly throw at PACS. And it is time for radiology to rediscover its soul. Radiology needs to answer the following questions at a minimum: What do we do? How do we do it? Why do we do it? And how can we do it better? What value do we bring our referring colleagues, our health-care system, our payers, our patients?
And just as Nadella posits that AI is the tool that will help Microsoft rediscover its soul, it is also the tool that will help us rediscover radiology’s soul. In the short term, we can rapidly develop several solutions—low-hanging fruit—that can easily help us manage the data explosion: tools to maintain our worklists and help us prioritize studies that need our attention sooner; tools to help us protocol our studies so that we perform the right test on the right person at the right time; tools that will take out the drudgery in our jobs by quantitatively measuring the size of masses on sequential imaging studies. Our society will then use supervised learning where we will teach machines to replicate some of the tasks that we currently outperform machines. Subsequently, we will exploit unsupervised learning where machines will discover better ways of doing some of the things that we do as radiologists. But AI will not be able to replace us entirely even technically. Current AI efforts have achieved limited success in narrowly focused image interpretation problems, and there is no indication that an AI system can practice general radiology. That having been said, radiology needs to do some soul searching to redefine its role in healthcare given the increasingly prevelant role of computers. The advances in the future will free us to do what we should be doing more of, which is to reconnect with our patients and interpret test results for them. AI will allow us to approach what should be our true goal of providing patient- and family-centered care. More important than technical capabilities, AI will never replace us because we can and will provide this care with empathy!
It is very heartening to see our field has welcomed this tool with open arms. For the last 15 years that I have been attending the Radiological Society of North America (RSNA) annual meeting, I have been fascinated by the various disparate advances that are made in our various subspecialties. The last annual meeting had a distinctly different feel to it—all subspecialties are now converging on AI and its current incarnation, deep learning. This convergence is providing the critical mass that is required for any breakthrough. With such a coordinated effort, we are likely to either succeed or fail, although I doubt the latter, spectacularly and quickly. Within the Association of University Radiologists, the Radiology Research Alliance formed a task force last year, whose deliberations have resulted in a forthcoming white paper in Academic Radiology . Subsequently, the Radiology Research Alliance has formed a working group to continually study the impact AI will have on our profession. RSNA is initiating a new journal, Radiology: Artificial Intelligence , with an inaugural issue to be published later in 2018. The Journal of the American College of Radiology recently published a special issue on AI in radiology ( e.g., Ref. ). The American College of Radiology has an AI Advisory Group. These are but a few of the organized responses from our society to AI in radiology.
The article by Nawrocki et al. in this issue, is both timely and informative . The authors provide a beginner’s introduction to AI and deep learning, and transition rapidly to provide insights into the technical challenges in developing, and societal challenges in adopting, AI advances in radiology. Equally important, they discuss the problems in radiology that can be readily addressed with smart investment in AI. We all owe it to ourselves to learn and prepare for the wave that is approaching our shores. It is up to us whether we let this wave turn into a disastrous tsunami or we ride this wave to reach greater heights. And in the process, we will hopefully rediscover radiology’s soul.
References
1. Nadella S.: Hit refresh: the quest to rediscover Microsoft’s soul and imagine a better future for everyone.2017.Harper CollinsNew York, NY
2. McBee M.P., Awan O.A., Colucci A.T., et. al.: Deep learning in radiology. Acad Radiol 2018; (in press)
3. Thrall J.H., Li X., Li Q., et. al.: Artificial intelligence and machine learning in radiology: opportunities, challenges, pitfalls, and criteria for success. J Am Coll Radiol 2018; https://doi.org/10.1016/j.jacr.2017.12.026 Epub ahead of print
4. Nawrocki T., Maldjian P.D., Slasky S.I., et. al.: Artificial intelligence and radiology: have rumors of the radiologist’s demise been greatly exaggerated?. Acad Radiol 2018; 25: pp. 967-972.