Deepfake X-Rays Are Fooling Radiologists — And AI Can't Stop It

Hey there, welcome to Beyond the Scan! I'm Jackie, your host, and I'm so glad you're joining me today, Tuesday, April 7th, 2026.

Quick shoutout to our sponsor RadiologyJobs.com — they're doing amazing work connecting imaging professionals with great opportunities nationwide.

I've got some really exciting stories to share with you today!

Alright, let's kick things off with something that honestly kept me up a little last night thinking about it. <break time="0.8s" /> What if a chest X-ray sitting in your PACS right now... wasn't real? I mean, what if it was completely fabricated — a deepfake — and neither you nor your AI tool could tell the difference? That's not a sci-fi scenario anymore. According to a study highlighted by RSNA News, deepfake chest X-rays have been shown to deceive both radiologists AND AI diagnostic models. We're talking about synthetically generated images that are convincing enough to slip past experienced human eyes and sophisticated software alike. And the implications here go way beyond just a cool tech demo — this is a genuine cybersecurity concern for imaging workflows and diagnostic trust. Experts are now calling for better detection methods, because if we can't verify the authenticity of an image, the entire foundation of image-based diagnosis gets a little shaky. I think this is one of those stories where our field needs to get ahead of the curve before it becomes a real patient safety issue. Something worth bringing up at your next department meeting, honestly. The question I keep coming back to is — how do we build verification into our systems before this becomes a widespread problem?

Now, switching gears to something that's going to feel a lot more immediately satisfying — because this next story is about making your workday better, right now. <break time="0.7s" /> A hospital has implemented AI software on its MRI scanner and cut exam times by more than fifty percent. More than HALF. That's not a typo. According to Radiology Business, the technology is boosting throughput without compromising image quality, and radiology leaders at that facility are reporting real improvements in both patient experience and overall department efficiency. Think about what that means in practice — patients spending less time in the bore, which is huge for anyone with claustrophobia or pain management challenges. And from an operational standpoint? You're essentially doubling your scanner's capacity without buying a second machine. I love this one because it's a perfect example of AI doing what it does best — not replacing anyone, but removing friction so the whole team can do more. If your department is still wrestling with MRI wait times and scheduling backlogs, this is the kind of development worth tracking closely.

Speaking of things that affect the whole department — let me tell you about what's happening on the policy front, because this stuff matters more than people sometimes realize. On April 2nd, the American College of Radiology was up on Capitol Hill for what they're calling the Future of Health Hill Day, advocating alongside other societies for federal support of medical imaging. The conversations centered on reimbursement and making sure imaging professionals have access to the technology they need to do their jobs. Now, I know policy talk can feel a little removed from daily clinical life, but here's the thing — reimbursement rates directly affect staffing, equipment budgets, and whether your department can even afford the AI tools we keep talking about on this show. The ACR showing up and making that case to lawmakers? That's your profession being represented at the highest level. <break time="0.6s" /> And if you've never thought about getting involved in advocacy yourself, this is a good reminder that those conversations are happening right now, with or without your voice.

Alright, this next one — you're going to want to hear this, especially if you work in breast imaging. A real-world study — and I want to emphasize real-world, not a controlled lab setting — looked at nine experienced breast radiologists reviewing over one hundred thousand digital breast tomosynthesis exams across four different sites. And when AI was added to their workflow, cancer detection rates went up. Here's the part that really stands out to me: sensitivity improved without increasing false positives. That's the balance that's always been so tricky in mammography — catching more cancers without sending more patients down a stressful and unnecessary biopsy path. According to Radiology Business, the findings from this study support broader AI integration in mammography. I think what makes this so compelling is the scale — a hundred thousand exams, four sites, nine experienced radiologists. This isn't a small pilot. This is real evidence building a real case. For anyone in breast imaging wondering whether AI is worth the investment, studies like this are starting to answer that question pretty clearly.

And finally, let's talk about something that hits close to home for a lot of folks in our community — workforce and the future of radiology subspecialties. Data from the National Resident Matching Program shows that emergency radiology is among eight subspecialties that failed to fill even fifty percent of their fellowship positions. <break time="0.9s" /> That's a significant gap. Meanwhile, neuroradiology filled eighty-seven percent of its spots, which tells us there are real preferences at play when residents are choosing their paths. Now, emergency radiology is such a critical area — it's the backbone of after-hours coverage, trauma response, rapid turnaround reads. A shortage of fellowship-trained emergency radiologists has ripple effects across entire health systems. My take on this is that we need to be asking why — is it the work schedule? Compensation? Lack of exposure during residency? Because understanding the why is how we start to fix it. If you're a program director, a department chief, or even a resident thinking about your future, this data is worth sitting with.

Well, that's all we have for today's episode of Beyond the Scan. I've really enjoyed sharing these developments with you.

If you found today's discussion helpful, I'd love it if you'd subscribe and share with your colleagues. And hey, swing by our website at beyondthescan.io where we've got some great additional resources on these topics.

I'll be back next week with more exciting updates. This is Jackie signing off, and remember — what we do really does make a difference in people's lives every day. Take care!