Overview

• The AI system was tested in real time across 12 Northwestern Medicine hospitals, analyzing nearly 24,000 reports over five months and delivering a 15.5% average lift in report completion speed, with some radiologists achieving up to 40% gains.
• By generating reports that are about 95% complete and personalized to patient data, the model lets radiologists review and finalize findings without compromising clinical accuracy.
• Integrated triage functions flag urgent conditions such as pneumothorax in milliseconds, helping emergency radiologists prioritize critical cases faster.
• Developed in-house on Northwestern Medicine clinical datasets, the lightweight model outperforms generalized AI tools while requiring less computational power.
• Now under two granted patents with more pending, the technology is entering commercialization, and follow-up work indicates it could boost efficiency by up to 80% and extend to CT imaging.