In a groundbreaking development, researchers at Northwestern University have unveiled a powerful generative AI tool that significantly improves radiologist efficiency—marking a new era in clinical imaging. Published today in JAMA Network Open, the study shows the AI-driven system streamlines radiology workflows, delivering up to 40% faster report generation without compromising diagnostic accuracy.
“To my knowledge, this is the first time AI has truly improved productivity in healthcare at this scale,” said Dr. Mozziyar Etemadi, senior author and assistant professor at Northwestern University Feinberg School of Medicine and McCormick School of Engineering. “Even outside of medicine, I’ve never seen efficiency gains like this.”
AI That Writes Radiology Reports and Flags Critical Conditions
Unlike existing tools that are narrowly focused on detecting a single disease or imaging type, Northwestern’s AI model is trained to handle all types of X-ray images. Drawing on data from radiographs and electronic health records—including patient history, exam reasons, and procedure types—the system generates radiology reports that are 95% complete. Radiologists then simply review, finalize, and approve them for clinical use.
Over a five-month deployment across 12 hospitals, the AI system analyzed nearly 12,000 radiographs, improving radiologist documentation efficiency by an average of 15.5%, with top-performing users reporting up to 40% faster workflow.
“For my team, it practically doubled our productivity,” said Dr. Samir Abboud, Chief of Emergency Radiology at Northwestern Medicine. “This AI tool is a force multiplier in every sense.”
Improving Patient Safety Through AI-Powered Triage
Beyond productivity, the study highlights another critical use: flagging life-threatening conditions. The AI system was particularly effective in detecting pneumothorax (collapsed lung)—a condition that can rapidly become fatal if missed. In several instances, the tool caught cases that had been delayed or overlooked during busy emergency shifts.
“In the ER, we’re often staring at a hundred images without knowing which one hides a critical diagnosis,” said Abboud. “This technology allows us to triage faster and treat sooner—potentially saving lives.”
Tackling the Global Radiologist Shortage with Smarter Tools
As clinical imaging demands grow and radiologist shortages worsen worldwide, Northwestern’s generative AI tool offers a practical solution. During the study, the efficiency boost translated into 12 fewer radiologist shifts needed over five months to maintain full hospital coverage—highlighting the potential for burnout reduction and resource optimization.
Importantly, the AI is not meant to replace radiologists but to support them.
“Medicine evolves constantly. A radiologist will always be the gold standard,” Abboud emphasized. “Our role now includes overseeing AI outputs and ensuring every report is right for each patient.”
Custom-Built, Low-Cost, and High-Performance
Unlike many off-the-shelf AI tools, Northwestern’s model was custom-built from scratch using its own clinical data, making it faster, more accurate, and less dependent on high-end computing resources. This approach ensures full transparency and adaptability while avoiding the pitfalls of commercial tools like ChatGPT.
“Our work proves that health systems can build their own AI tools without relying on expensive black-box solutions,” said Etemadi. “This is a step toward democratizing AI in healthcare worldwide.”
What’s Next: CT Scans and 80% Efficiency Gains
The research team is already testing a second-generation version of the tool in an unpublished follow-up study, reporting efficiency gains of up to 80% and expanding its use to CT scans. If validated, this could further revolutionize diagnostic imaging across clinical specialties.
Key Takeaways for Healthcare Leaders:
-
Efficiency Boost: Up to 40% increase in report generation speed.
-
Clinical Accuracy Maintained: Peer-reviewed validation on 800 patient cases.
-
AI-Assisted Triage: Improved detection and prioritization of critical conditions.
-
Scalable and Cost-Effective: Built using existing hospital data and infrastructure.
-
Supports Radiologists: Not a replacement, but a powerful tool for modern care delivery.
As the first generative AI radiology tool integrated into real-time hospital workflows, Northwestern’s innovation represents a turning point for AI in medicine—transforming radiology into a faster, smarter, and more patient-focused practice.
