AI Insights: A Tighter Bond
By Angela Adams
Radiology Today
Vol. 26 No. 7 P. 26

AI won’t replace clinicians, but it could fail patients if we let it.

As hyperbolic headlines alternate between warning about the coming AI apocalypse and heralding a coming AI utopia, it’s worth remembering that we’ve been here before. One of my favorite examples comes from 2017, when a renowned venture capitalist emphatically declared that “the role of the radiologist will be obsolete in five years.” He wasn’t alone. Some other estimates put the pending obsolescence at 10 years. While it is clear that radiologists are not obsolete and, in fact, are in short supply, as noted in the November/ December 2023 issue of Radiology Today, market bullhorns still claim that the end of the radiologist is in sight.

As attention grabbing as these claims are, they have shaped the AI and health care conversation, and the shock value has had lasting impacts. A recent survey from the Pew Research Center reports that more than half of US workers are worried about AI. What has happened within radiology in the almost decade since the predicted “death of the radiologist” can serve as a guide to those in health care and beyond to help navigate AI’s promises, pitfalls, and patient impacts.

The lesson? To make the most of the AI opportunity immediately in front of us, we need to reframe the conversation from one about replacement to one about patient outcomes.

From my years as a bedside nurse to my current role leading a technology company, I’ve seen how health care really works: nothing happens in isolation. A finding on a scan means nothing if it never reaches the right provider. A recommendation for follow up doesn’t save lives if no one makes sure the patient actually receives it.

AI can—and does—outperform humans at certain technical tasks. And we are learning a tremendous amount about the relationship between AI and humans, including what it takes to embed AI into clinical workflows safely, ethically, and with the human at the helm. But we have to remember that health care isn’t a technical task. It’s a chain of trust, relationships, and accountability. When technology disrupts that chain without closing it, patients fall through the cracks.

First, a Little History
Medical imaging has been undergoing a digital transformation for the past 50 years and is now becoming ubiquitous. It comprises one of the largest, if not the largest, sets of data in health care, and the rate of imaging data generation continues to exponentially increase. Rates are compounded by the rise of multiscreening applications, the increase in disease burden, an aging population, and accessibility of imaging technologies for preventive care.

The job of a radiologist is to help referring clinicians and patients make sense of imaging data in the context of their specific clinical scenario. Radiologists have deep subject expertise cultivated over many years of intensive training and continued refinement—and that goes far beyond mere image interpretation. Image interpretation includes classification, measurement, and reporting—all tasks that AI is capable of—but the most valuable part of the radiologist’s role comes through what happens next.

Crucially, a radiologist can place that finding in context, making it actionable for each specific case. As medical imaging data continues to proliferate, the demand for that skill set becomes greater.

A Flood of Data
Think of data as a reservoir, with a river of raw information coming in one end. On the other end, a dam: data processing and output. Historically, data generation has come primarily from image acquisition, and the outlet has been radiologists processing the data.

Advanced imaging technologies and AI-assisted tools generate more data per scan than ever before. These tools identify more findings—many incidental, some critical. The result? A deluge of data and more follow-ups to manage. The math leads to a straightforward result: We need more radiologists. We’re facing an increased need for systems as well as people who can make sense of AI-enhanced findings in a high-stakes, time-sensitive environment.

Just because a technology may be invented that can perform some of the functions of a role does not necessarily mean that it will make the practitioners of that specialty obsolete. Simply layering automation onto current systems is insufficient; clinicians and leaders must intentionally define tasks and protocols so that AI fits into discrete parts of care delivery. Errors and inefficiencies that happen in the space between the AI and humans will persist unless workflows are fundamentally redesigned around AI. But integration isn’t the goal. Better patient outcomes continue to be, and will always remain, the focus and function of health care.

Like radiology, many areas of medicine are seeing an explosion of data. AI is surfacing more incidental findings than ever before. But here’s the part that rarely makes headlines: If health systems don’t have the infrastructure to act on those findings, all AI does is speed up the rate at which critical care opportunities are missed.

This isn’t just a radiology problem. It’s an oncology problem, a cardiology problem, a primary care problem. And it’s a problem that technology alone cannot solve.

The Right Conversation
Like most clinicians, I’m not interested in clinging to the past. I’m interested in delivering on the promises we make to our patients every day. In my years as a nurse, I learned that the moment of diagnosis is only the beginning. What matters most is what happens next. Did the patient get the follow-up imaging, the specialist appointment, the treatment plan they needed? If the answer is no, then every algorithm and data point loses its value.

The “Will AI replace you?” debate misses that reality. It’s not about turf. It’s about trust.

Technology should strengthen the chain from finding to follow-up, not introduce more weak links. That means embedding AI into workflows in ways that make it easier for clinicians to do their jobs reliably, without losing sight of the human relationships and accountability that define good care.

We need to move beyond disruption for disruption’s sake. Health care is complex, high-risk, and deeply personal. That’s exactly why improvement demands humility, partnership, and a relentless focus on outcomes.

So, to those who predicted the “death of the radiologist” or the “end of the clinician,” the conversation needs to shift. We need to talk about outcomes and experience, and make sure every patient who needs follow-up gets it, every time. Because, in the end, AI isn’t replacing the clinician. It’s challenging us to do what we’ve always done best: show up for patients and see the work through to the best possible outcome for every individual, every time.

— Angela Adams is the CEO of Inflo Health.