Flash Versus Function: Does the AI Focus in Radiology Compromise Our Infrastructure Needs?

Addressing upgrades of traditional infrastructure used in everyday radiology practice may be a more practical use of resources than investment in artificial intelligence (AI) technology that is still evolving.

I recently got an unexpected series of emails in regard to a comment I made on a Twitter thread. The thread had begun with a lament about radiology departments directing their efforts and budget to newfangled things like artificial intelligence (AI) rather than traditional elements of their infrastructure (PACS, radiology information system (RIS), dictation software).

I had commented that, unfortunately, it’s not as flashy to boast “We have the same type of software products as everybody else, but our products are quality” as it is to brag “We have something cutting-edge that nobody else does,” even if the cutting-edge stuff doesn’t get the job done. I added that the sad truth of the matter is people who have authority to allocate resources are often more motivated by flash than function.

For a typical radiologist’s perspective and many non-rads, this seems a no-brainer. Why would valuable resources be thrown after gimmicky new fripperies that might not even work instead of applying them to tools that are routinely used to make ends meet?

For instance, suppose I have a clunker of a car that routinely fails to start, stalls out or otherwise dies midway to my destination, not to mention getting lousy mileage whenever it does manage to work. Shouldn’t I make fixing it a priority or replacing the whole thing with something more reliable and efficient? Wouldn’t this be a better approach than tricking it out with a new paint job, spoilers, and ground effects?

Now suppose I depend on that car to make my living. Perhaps I need it to commute to my job or perhaps it is central to my job (such as driving for Uber). Making the vehicle more reliable enables me to keep paying my bills. Fixing/replacing it will not only “pay for itself,” it may improve my bottom line as I subsequently need fewer emergency repairs, get more gigs done and/or bonuses for actually showing up on time, etc.

None of this will happen if I instead choose to detail my vehicle. Guys on the street corner might be impressed as I drive by (assuming I don’t break down in front of them), but they’re not the ones paying my bills or talking to my boss, clients, etc. about what a cool driver I must be.

Even if someone related to my livelihood sees my tricked-out vehicle, he or she might not be impressed with my choices. Instead, if that person has heard I often cite car trouble as a reason I fail to show up, he or she might decide I’m not a smart or reliable person. Alternately, that person may resent having to cover for me at times and here I am blatantly doing nothing to improve the situation.

Seeing someone make such seemingly illogical choices, I always try to remember there are any number of things I might not know or understand about their situation. There could be a reasonable explanation that would have me nod and think, “Okay, now that makes sense.”

Without that intel, though, I’m liable to think that the individual either doesn’t understand things well enough to make better decisions or doesn’t care to make better decisions. My conclusion will surely be colored by the details of the situation. I am unlikely to assume a teenager who tricks out his clunker car has a wise rationale. However, if a CEO does something that makes no sense to me, I have to imagine he or she may have valid reasons.

Still, someone in command of a radiology group who decides to throw resources at AI instead of PACS, RIS, voice recognition, etc. is far from immune to scrutiny. This is especially the case if the latter stuff—the tools that literally make it possible for the rads to pay the bills and keep the group afloat—is constantly breaking down, working at a fraction of the speed/efficiency it should be, contributing to lower quality work, or burning out the rads who have to live with it.

Unless that AI is somehow going to counterbalance the bargain basement nature of the other infrastructure (or it magically comes with a reliable, separate stream of revenue to pay for itself), it sure looks like whoever decided to get that AI either doesn’t understand or doesn’t care how badly the existing infrastructure is hurting the group.

That can easily happen when the decision-maker is someone who doesn’t work “in the trenches” with the other radiologists. Someone who doesn’t regularly experience the inefficiencies and other flaws in the infrastructure might not be all that motivated to fix them. She or he might not even get that there is anything to fix.

At the very least, a leader who doesn’t put in time on the worklists but listens to and respects working rads would have a chance of understanding what needs fixing and/or replacing. I emphasize that second part because it’s all too common to be dismissive: “Oh, you know radiologists, they always complain about X.”

Such an understanding paves the way to appreciating that improving the rads’ tools can boost overall productivity, quality of work, and team morale by a staggering amount. This more than pays for itself and maybe helps a rad group afford extras like AI in future years when the technology will unquestionably be better than it currently is and likely cheaper as well.

Take away that firsthand (or close secondhand) comprehension of the state of affairs, and it’s not hard to imagine why leadership might leave a crummy infrastructure in place. Why throw more resources at new PACS, RIS, etc. when, as far as you can tell, the systems you already have seem to get the job done? Besides, everybody has that stuff. Let’s go get ourselves some sort of AI instead and be the first rad group in town with one!