AI and the future of Radiology

Artificial Intelligence applications in Radiology have gotten a lot of buzz lately, mostly due to advances that have made the industry stand up and take notice. These advances have led to the very genuine and even legitimate concern that it could impact radiology on a scale that could threaten your job. We’ll cut straight to the spoiler on this one…Don’t start looking into Plumber School just yet. While it’s true that AI has become pretty impressive at identifying pathologies such as lesions and fractures, it’s performance is about the same as a human, which is pretty impressive to be sure, but is it practical for an AI radiologist to be a substitute to you? We don’t think so.

It’s very likely that AI will play an integral role as a part of the radiologists toolbox, mostly because it has the ability to review multiple, even hundreds of images quickly and let’s face it, images are multiplying a lot faster than Radiologists. Anything that reduces labor costs while simultaneously improving diagnostic accuracy is a benefit to both physicians and patients that no one is likely to argue about, however, a radiologist is not just a reader of images, consider that before you start reassessing your career choice. Deep learning AI models need to be programmed to perform hundreds, even thousands of single tasks in order to be able to have a comprehensive view, in other words, suggesting that they are “intelligent” is a bit of a misnomer, they aren’t “thinking” as much as they are looking at the multiple recognition tasks we teach them to identify. In other words, if we only had them looking for a single task, like nodule detection, they wouldn’t “see” anything else. Currently AI only has the ability to execute a handful of these tasks. Additionally, a “real live” Radiologist consults with other physicians, an AI only has a narrow or binary perspective, and of course does not have the benefit of experience or additional information like the patients personal or family medical history to factor into it’s interpretations. Probably the biggest reason for you to be encouraged that Hal 2000 is not stealing your job is that AI is a Diagnostic tool, and offers nothing when it comes to treatment or actual diagnosis.

Even in the unlikely event AI became the standard for reading, even interpretation, there’s still plenty of other things for you to do, besides, no patient is interested in their Radiologist being replaced by a Red Box like Blockbuster Video was. Patient interaction and the actual care that a patient expects when confronted with serious health conditions are not likely to ever be replaced by “smart” machines.


0 Responses on AI and the future of Radiology"

© Copyright 2015-19 Rock the Boards, L.L.C. | | Privacy Policy