As much as the relationship between artificial intelligence (AI) and radiology has already developed, it is still in its earliest stages. What will that relationship look like in a decade? Or in another 20 or 30 years?
Bernard F. King Jr, MD, department of radiology at Mayo College of Medicine in Rochester, Minnesota, wrote at length about the future of AI and radiology in a new opinion piece for the Journal of the American College of Radiology.
“In this perfect storm of rapidly developing deep learning algorithms and artificial neural networks, along with the explosion of big data and the acceleration of processing power, we have witnessed the beginning of a new world of AI,” King wrote.
These are four key predictions from his analysis:
1. AI’s impact in radiology will take time.
King explained that AI isn’t going to completely overhaul radiology overnight; it’s going to take some time. The first stage of this change is already here, he added, and can be see in AI’s ability to perform “automatic segmentation of various structures on our digital CT or MR images.”
“Segmentation of structures is the first step in any effort to isolate and analyze organs or pathologic lesions for analysis,” King wrote. “Although segmentation of structures seems easy and readily apparent to human operators, it can take enormous amounts (hours) of time to perform by humans. To date, it has been very difficult to automate segmentation with conventional computer programs. Fortunately, AI systems have already been developed to solve this conundrum.”
This is an example of AI’s ability to give radiologists more time to demonstrate their value by working on bigger, harder problems. Upcoming stages of AI’s growth in radiology will likely involve more complex artificial neural networks and deep learning algorithms, King added.
2. The accuracy of radiologists’ diagnoses will improve significantly.
Healthcare is more focused on value-based care now than ever before, and radiology will certainly be able to improve the value it provides to patients thanks to advancements in AI technologies.
“AI will open new inroads into radiology diagnosis that have heretofore been impossible through mere image interpretation,” King wrote. “Image texture analysis combined with pathologic and genetic correlation will allow AI systems to learn from vast amounts of data and help improve the accuracy and value of our diagnoses.”
3. Economic factors will fuel the growth of AI.
King pointed to one reason the influence of AI will continue to spread in radiology: Businesses have a chance to make a lot of money.
“Health systems and payers are constantly looking for higher levels of efficiency and lower costs,” he wrote. “This drive for more efficient and lower cost of care will definitely play a role in future AI applications in medicine and radiology. Entrepreneurs and venture capitalists are also constantly looking for new discoveries and applications in the medical world.”
4. AI won’t take jobs from radiologists—as long as radiologists can adapt.
King doesn’t necessarily think radiologists will be replaced by AI applications and algorithms—but the specialty will have to adapt quite a bit as time goes on.
“As many of us know, the medical profession of radiology is much more complex and demanding than what AI will be able to do in the future,” King wrote. “Therefore, the human intelligence of a radiologist will always be at the forefront of diagnostic imaging and patient care, but our roles as experts in image-based diagnosis and intervention will likely change over the coming years. If we, as radiologists, do not adapt to this change, we may become extinct.”