At RSNA 2025, two radiologists dismantled the 2016 prediction that AI would soon make the profession obsolete, instead describing a future where AI tools are critical allies in a workforce stretched thin by rising imaging volumes and an aging radiologist population. With turnover rates in the United States climbing from 5.3% in 2013 to 8.5% in 2022 and roughly 19% of European radiologists expected to retire within five years, the emphasis has shifted from replacement to augmentation.
The workforce crisis behind the shift
Nancy Pham, MD, Assistant Professor of Radiology at Stanford Medicine, pointed to an unsustainable gap between exam volumes and the number of available radiologists. "A large and unsustainable imaging volume is outpacing the global baseline radiologist supply," she said. The growing pressure is driving radiologists out of the profession at accelerating rates, even as demand for diagnostic imaging continues to rise.
Three types of AI entering clinical practice
Pham categorized AI tools into three groups, each with a distinct role in easing the workflow. Autonomous AI makes clinical decisions without human oversight. Two products are in use: LumineticsCore, the only FDA-cleared autonomous AI for detecting diabetic retinopathy, and the CE-marked Oxipit ChestLink, which identifies normal chest radiographs with 99.9% precision so radiologists can focus on abnormal cases. Generative AI tools can rapidly synthesize patient histories, free-text notes, and lab results into structured summaries. Pham uses one such tool at Stanford Medicine when patients have complex, lengthy records. AI Agents & Automation handle repetitive, non-interpretive tasks like segmentation, volume measurement, and lesion tracking. Assistive AI also serves as a second reader, catching false positives and negatives-particularly in breast imaging, where the evidence is strongest.
Noncontrast head CTs: the high-stakes automation target
"Noncontrast head CT scans are one of the highest volume exams ordered by emergency department physicians," Pham said. "Often performed daily for minor trauma, headache, or dizziness, the vast majority of these CT exams are normal." An AI tool that screens normal studies could free radiologists to concentrate on complex cases. She added a warning: "The safety bars for such a tool are incredibly high. The brain is unforgiving."
The highly automated reading room of tomorrow
Tara Retson, MD, PhD, Deputy Chief of AI in Breast Imaging at UC San Diego Health, predicted that reading rooms will soon integrate multiple image-specific AI tools with PACS and electronic medical records. These systems will extract information in seconds, generate summaries of prior studies, and provide real-time alerts, triage, and prioritization of unread exams. AI for Healthcare will also standardize BI-RADS reports, do real-time coding, and embed guidelines directly into reports.
"I believe that the biggest impact AI will have will be automating tasks, not performing activities that will replace radiologists," Retson said. She emphasized that AI will promote patient-centered care and allow radiologists to focus on complex imaging by removing tedious, time-consuming work. Still, she urged caution: "AI fails in ways that we don't expect. No system is perfect. AI hallucinations are real."
Why this matters for healthcare professionals
For radiologists and imaging department leaders, the message from RSNA 2025 is clear: AI adoption is no longer about preparing for a distant future. It is a practical response to immediate staffing shortages and burnout. The tools that automate routine reads and standardize reporting will change daily workflows, but human oversight remains non-negotiable. The focus for teams should be on rigorous quality control, transparency, and selecting AI systems that genuinely reduce the burden-not add noise.
Nancy Pham, MD is an Assistant Professor of Radiology in neuroimaging and neurointervention at Stanford Medicine.
Tara Retson, MD, PhD is an Assistant Professor of Radiology at UC San Diego Health, focusing on deep learning applications in medical imaging.
Your membership also unlocks:
