JACR® Associate Editor for Health Services Research Gelareh Sadigh, MD, talks to the Bulletin about what readers can expect from the journal’s latest focus issue.
Interview by Alex Utano, associate editor, ACR Press
The times are changing across all professional landscapes as technology continues to advance. Rapid changes in what tools are available to help us do our jobs have prompted many of us to adapt to the times or risk being left behind. The use of AI is a common factor throughout this evolution, and as AI technologies evolve, more institutions across the world have adopted them into their daily workflow. Radiology is no exception.
The JACR’s March Focus Issue on AI on Workflow Optimization offers a collection of articles that explore the ways in which AI is being regularly utilized across practice types. The Bulletin spoke with Gelareh Sadigh, MD, associate editor for health services research, about the creation of this timely focus issue and what readers can expect to learn.
The main thing that drew us to this topic was how different the promise of AI sounds compared to how it actually shows up in our day-to-day work. We hear a lot about AI accuracy and innovation, but in practice the question is usually, ‘Does this AI tool save me time or does it make my work more efficient or more sustainable?’ In my opinion, workflow is where AI either proves its value or fails. If using AI slows people down, adds more clicks or increases their cognitive load, then it won't last. So, no matter how good the model is, from a health services perspective, AI is really a system intervention and not a technical one. That’s the lens that my co-guest editors Brian W. Bresnahan, PhD, Nathan Cross, MD, MS, CIIP, Jonthan Medverd, MD, and I wanted this focus issue to take.
We designed this issue to be intentionally broad but practical. There’s some original research on topics like deep learning for head CT, large language models to improve radiology report readability and using AI to track follow-up recommendations. These are problems that radiologists deal with on a day-to-day basis. But also, this issue includes reviews and frameworks that step back and ask bigger questions: How do we deploy AI tools in a way that actually sticks? How do we link workflow assistance to real outcomes? How do AI business cases support learning health systems? We also have some opinion pieces on AI governance and legal considerations because workflow optimization doesn't just happen in a vacuum. It usually happens in a real institution where you have real limitations and constraints.
Whenever a new AI technology comes in, I now think to myself ‘Who would benefit from this AI tool? Who might be burdened by its unintended consequences?’
There were a couple of things. One is that I was surprised at how often we usually talk about workflow impact, but we don’t measure it in our studies. We have many AI tools that have a reputation of improving our workflow, but there are limited studies that quantify things like how much time-saving these AI tools bring us and what the downstream effect is on patient care. The other thing that surprised me was how context-dependent the success of AI tools is. The same AI tools can help one practice but may fail in another based on how it is integrated into their clinical workflow, how their staffing or local culture is. That reinforces the idea that AI is not just a plug and play. You need actual implementation in every practice. I was also encouraged to see that a lot of authors had interest in writing about AI governance and real-world implementation, which shows the field is maturing in AI.
It made me more intentional about asking where AI fits into our day-to-day workflow, not just whether or not it works. Whenever a new AI technology comes in, I now think to myself “Who would benefit from this AI tool? Who might be burdened by its unintended consequences?” It also reinforces how critical interdisciplinary collaboration is. We as radiologists alone cannot optimize workflow. We need AI technologies, informaticians and administrators. Sometimes we need a patient at the table, and AI works best when it’s designed with a full system in mind.
I think workflow isn’t usually a secondary benefit of AI. It’s the main determinant of whether a tool is successful or not. If AI is going to meaningfully help radiology, it must make care delivery better and not more complicated.
The focus issue basically reflects a broader shift in radiology. We are moving from asking, “Can we build AI tools?” to asking, “Does this AI tool have value? And who does it have value for?” And if we take that lens, I think AI really has the potential to improve radiology workflows and healthcare delivery.
How Repeat Participation in the ImPower Program Drives Lasting Change
For institutions that consider joining or returning to the program, repeat participation reinforces that quality improvement is not a one-time initiative but an ongoing process.
Read more
The Return of Oral Boards
As the ABR rolls out updates to its certifying process, trainees should know what to expect as they navigate these changes.
Read more
Reflections on What We’ve Accomplished Together
Through a global pandemic, pressure on the Medicare conversion factor, the emergence of AI and more, the ACR Commission on Economics did not just react — it led.
Read more
