The new kid on the block in medical imaging is the exciting new technology of CAD or computer aided detection. The essence of this technology is its ability to use computer math algorithms to
analyse radiology images. Given that radiologic images are today essentially digital ie, a combination of bits and bytes (as is, coincidentally, the title of this blog) analysis of the digital code that constitutes the medical image can be used to detect imaging patterns, in a manner akin to the interpretation of the DNA helix sequence to detect genetic diseases.
The question that I get asked on a regular basis is “when will the computer replace the radiologist?” The answer I give to that question is “not anytime soon”
This response is not intended to downplay the importance of CAD technology in any way. CAD is a truly transformational tool, akin to solar powered cars or the internet in terms of the way it can potentially impact on healthcare quality, efficiency and costs.
My point is related to my belief that radiology is both an art and a science. There is a left brain and a right brain component to radiologic analysis that are both equally critical and irreplaceable.
In that sense, CAD is a valuable support tool for radiologists, but not a replacement for them. CAD plays the role of the support team (fitness trainer, physio, coach and nutritionist) for Novak Djokovic or the rest of FCB for Lionel Messi’s goal scoring skills. However, could you imaging Novak’s fitness trainer playing at Roland Garros in his place? Or the left back of FCB taking over Messi’s role as chief goal scorer in Barcelona’s Poble Nou Stadium?
While I appreciate that the analogies are crude, I would submit that the role of the radiologist in medical imaging analysis today needs to be viewed in the same light as that of a key striker in a football or tennis game who can only perform at key performance levels if the rest of the process and team are in place. And skilled radiologists today while perhaps not as rare as a Djokovic or a Messi are certainly a rare breed, as all recent articles on radiologist burnout attest.
The fact is that many of the tasks that a radiologist performs in the course of analyzing a medical image are quite mechanical. For instance the detection of a nodule in the lung that may be an early cancer is a very very important task which was till date fulfilled by the radiologist however, today a computer algorithm can achieve the same results with equal efficiency
Similarly the detection of early breast cancers on mammography, or the detection of pulmonary emboli on chest CT or the detection of colon polyps on CT colonography are all techniques in which computers today equal, closely approximate or in some cases may even surpass the skills of radiologists. It is important to realize that radiology has become so diverse today that no single radiologist can possibly hope to do justice to its entire gamut and spectrum.
The part of the job where radiologists are irreplaceable is in the analysis of the patterns and their correlation with the clinical context, to create a consolidated assessment which ultimately is essential for patient care. It is the cumulative knowledge and awareness gained over 5 or 6 years of medical schooling. And in the succinct communication of such analysis to the patient’s treating physician, effectively conveying the degree of certainty or equally important the degree of doubt, about the significance of a radiologic finding. This is an even more complex skill which is initially gained in a rudimentary form during radiology residency, but takes years or even decades of radiologic practice to hone. Which is why it is so hard, and takes so long to develop good radiologists.
And so, until such time as we have a computer that not only detects but also analyses and communicates like a human radiologist (Siri is perhaps the first step in that direction!), I think radiologists’ jobs are pretty secure, in the same manner that Djokovic or Messi’s job is secure, at least as long as they continue to perform at the levels they currently are.
The crux of the issue is whether with increasing workloads, radiologists can in fact continue to perform at peak. I think not. As in every other profession, radiologists too are susceptible to a higher error rate if their clinical practices become too busy, which given current trends of radiology utilization, is an all too common phenomenon.
And so a hybrid practice of physician and computer working in partnership is ultimately in the interest of optimal patient care. As I see it, is the future of radiology lies in greater acceptance of CAD technologies by radiologists, and the appropriate direction of technology by radiologists to meet their clinical needs, based on their clinical performance and lacunae therein. In the similar manner in which Djokovic’s performance directs his offseason training needs.
The use of technology in sport has today allowed talented athletes to optimize their performance. In medicine, we can learn from these examples and use them to improve our healthcare delivery system to the benefit of both our patients and our profession.
The other core issues are sociopsychologic ones, a) that radiologists feel threatened by support systems that actually enhance their practice and that b) regulatory systems are structured in a manner that they support and promote the status quo and are suspicious of game-changing and transformational technologies, such as CAD represent.