Northwestern's Generative AI System Drafts Personalized Radiology Reports in Real Time
Northwestern Medicine researchers developed a first-of-its-kind generative AI that analyzes medical imaging and drafts personalized radiology reports in real time, boosting radiologist productivity by up to 40%.
Researchers at Northwestern Medicine have developed a generative AI system that represents a significant leap beyond current AI radiology tools. Rather than simply flagging findings, the system analyzes imaging studies and drafts complete, personalized radiology reports in real time — a capability that has been demonstrated to boost radiologist productivity by up to 40%.
The system integrates directly into the radiologist's reading workflow, generating structured draft reports that include findings, impressions, and clinical correlations. Radiologists then review, edit, and finalize the drafts, maintaining full authority over the final report while spending significantly less time on the mechanical aspects of report generation.
This approach addresses one of radiology's most persistent pain points: the documentation burden. Radiologists at busy academic centers may interpret hundreds of studies per day, and report generation consumes a significant fraction of their time. By automating the initial draft, the system allows radiologists to focus on the interpretive and clinical reasoning aspects of their work.
The research was presented at RSNA 2025 and has attracted attention from health systems exploring ways to address the growing gap between imaging volume and radiologist workforce capacity. The researchers noted that the system's accuracy improves with institution-specific training data, suggesting deployment strategies that customize the model to local reporting conventions.