How will artificial intelligence impact radiology? The answer you usually hear is: "It'll make us more efficient, improve workflows, and let us focus on the important tasks." Sounds great, doesn't it? But let's get real and ask the real question: Will AI replace radiologists?

The standard response is a resounding, "No, no, no! AI will only replace radiologists who don't use AI." The rest will be just fine.

In this article, I'm going to share my personal view, and it's probably less rosy than what you're used to hearing. As a radiologist and medical director of a practice, I see things from a slightly different perspective.

​

Let's be honest: What's the goal of any AI company developing these fancy programs? Easy – they want to make a lot of money! To do that, their product needs to be valuable to us, the radiologists, the hospitals, the practices. But what does "valuable" actually mean? Why would anyone in radiology buy this software?

The answer is always, ultimately, a financial benefit. Either the AI makes the radiologist more efficient (think voice recognition instead of typing), allows us to report more studies for the same salary, or provides some new insight that helps referring physicians (and acts as a marketing tool to get more referrals).

The AI company is selling an algorithm that, in the end, helps the radiology business make more money. Let's say a company creates an AI that checks your report and adds references to relevant literature. Sounds cool, right? But does it really add value, either to the busy radiologist or the even busier referring physician? Probably not. Now, imagine that AI costs $1 per report. The radiology practice can't charge the patient for that, so their reimbursement is now $1 less. The decision-maker will look at that and say, "Nice toy, but we don't need it." That company will go nowhere, despite good intentions.

​AI companies can charge more for their service the more money they save or make for the radiology department. A minor efficiency boost gets a small price tag. A major boost or cost saving? That's where they can really cash in.

So, when we analyze the AI boom's impact on radiology, it comes down to this one question: "Do we believe AI will become as good as a radiologist?" This is the question you need to ask yourself. And we're already seeing evidence that it's happening in certain areas, like breast imaging.

Take this study: Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol. 2023 Aug;24(8):936-944 by Lång K et al.

Conclusion of this study: Interpretation: AI-supported mammography screening resulted in a similar cancer detection rate compared with standard double reading, with a substantially lower screen-reading workload, indicating that the use of AI in mammography screening is safe. The trial was thus not halted and the primary endpoint of interval cancer rate will be assessed in 100 000 enrolled participants after 2-years of follow up.

That conclusion basically says that AI-supported mammography screening is just as good as standard double reading, but with a much lower workload. Technically, you might only need one radiologist plus AI, instead of two.

Now, breast radiologists will argue, "We do more than just screening! We do ultrasounds, MRIs, biopsies, counseling..." True. But let's say you have a team where two radiologists are dedicated to screening every day. That AI could potentially replace one full radiologist. A breast radiologist's salary in the US is, say, $350,000. So, if the AI company is smart, they can charge something like $100,000 per year for their program, and the practice still "saves" $250,000.

​

Here's another thing to understand: Who decides whether to buy AI software? It's not going to be a democratic vote among the radiologists. The leadership will look at the numbers, do the calculations I just did, and decide if it makes them money or saves them a significant amount. I know this because I'm one of those decision-makers. I've been pitched AI solutions. While the results might be nice, the decision for us is always a financial one. Can we charge insurance for it? In Switzerland, we can't. So, with reimbursement rates already dropping, why would we pay more for an analysis that cuts into our profit without a monetary benefit?

Do you believe AI can become as good as a radiologist? If yes, the only logical consequence is that we'll need fewer radiologists. Sure, you can argue there's a shortage now, and AI will help. Maybe in the short term. But the better the AI gets, the fewer radiologists will be needed. At some point, management will calculate whether they'd make more profit by adding an AI for $100,000 and giving notice to one of their staff radiologists, using the extra $250,000 to grow the business, boost the CEO's bonus, or pay dividends. It's a business, after all.

Can AI be as good as a radiologist? That's the only question. Medicolegal issues can be solved as long as there's one "last radiologist" willing to put their name on the AI-generated analysis (and do the ultrasounds, biopsies, and other hands-on tasks).

Look at OpenAI's new AI agent business model. They're selling PhD-level AI agents for $20,000 a month – $240,000 a year. These agents can do 24/7, 365 days a year, the high-level research you'd otherwise need a few PhDs for. Who would buy that? Any company that can save $240,000 in salaries with one agent.

I'm not an AI engineer. I don't know how far off this is. But every year, new AI companies try to enter the market. Image recognition is getting better, AI can create videos from text, and vice-versa. Radiology images are just data, and I'm convinced that, at some point, a company will emerge with an AI that can create reports at the same level or better than an average radiologist.

The radiology market is big enough to make serious money. It's not some niche area. Just put yourself in the shoes of a radiology group CEO. You have 10 radiologists, costing $3.5 million in salaries. You get offered an AI agent for $500,000 a year that can do the work of five radiologists, at the same or better quality. What do you think will happen? Your team shrinks, your profit jumps by around $1.25 million.

And here's another factor: Reimbursement rates will probably decline as a result of increased efficiency and profit. That means radiology groups that don't adopt AI will suffer first, either by cutting salaries or reducing staff and then adding AI.

So, in a way, the message "AI won't replace radiologists, but only those who don't use AI" is correct. But think about who's saying it. It's usually the chiefs of departments, the heads of corporations. From their perspective, it's true – AI won't replace all radiologists, but groups not using AI will be replaced by those that do.

​From the perspective of an employed radiologist, the future is less clear. But what can we do?

​Ask yourself, and ask your colleagues: Do you believe AI will become at least as good as an average radiologist? It's a simple yes or no question that will determine your future. Share your thoughts in the comments.