I recently read a piece that made a familiar argument: as disease complexity and aging populations exponentially increase our imaging backloads, it’s only a matter of when, not if, AI will begin to fully replace human radiologists. This development is inevitable; there are just too many studies and not enough people to go around to read them.

But this narrative that AI is disrupting radiology just hasn’t happened yet. We're putting the cart way in front of the horse. Will AI replace radiologists? Many radiologists don’t even feel AI yet, much less feel threatened by it.

As a radiologist myself, I’ve seen this firsthand. In fact, I know a radiologist who recently went to part-time and was offered $1 million-compensation package if they came back to full-time status. I don’t imagine radiology groups would be doing so if AI’s radiology applications were as far along and self-evident as articles often suggest it is.
Like a lot of radiologists, I don’t feel threatened by AI’s potential; I’m excited and eager for what it can do to help us work faster and more efficiently, bring down these imaging volumes we’re looking at every day, and accelerate patient care. I want AI in my job as soon as we can get it. And to be clear, AI, even in its limited use cases right now, does add some value to the table.

Consider this 2023 study published in Radiology that found up to 28% of chest x-ray interpretations could be safely automated by an AI tool with a sensitivity rate of 99.1%. A separate 2022 study published in Frontiers in Public Health likewise found AI was very useful at flagging lung nodules in CT scans.

Where AI can do the most obvious good for radiology
Despite the optimism that AI would revolutionize radiology, its impact to date has been much more modest than we anticipated. High implementation costs, coupled with studies showing that leading diagnostic tools perform only marginally better than radiologists, have tempered our expectations. Moreover, current AI solutions have not significantly improved physician time utilization like we may have hoped – or, frankly, are currently being talked about.

To deliver real value, the focus should shift toward efficiency-enhancing tools—particularly in areas like structured reporting, workflow automation, and decision support—to streamline processes, reap enormous time savings, and reduce radiologist burnout. Let’s look at a couple of examples:

Streamlining inefficient reporting
Building detailed and accurate radiology reports is an onerous, time-intensive process for radiologists who have little to no time to spare as it is. Measuring findings, tracking changes over time, and populating structured reports with those findings are repetitive efforts that can slow down radiologists and, worse, increase the risk of human error. AI tools can help reduce this burden by automating the integration of a radiologist’s findings into reporting.

Artificial Intelligence Homepage