Thursday, December 3 | 9:30 a.m.-10:30 a.m. | R3-SSIN07-1 | Room E450B
Researchers will present their experience in rolling out 11 triage AI algorithms across three hospitals.
Real-world evidence on large-scale AI deployment in clinical practice remains limited, noted presenter Domenico Mastrodicasa, MD, of the University of Washington, and colleagues. The group evaluated how system-wide implementation of multiple AI triage tools affected report turnaround time at a quaternary academic health system that includes a Level 1 trauma center.
The algorithms were deployed in stages between September 2023 and August 2025 and processed more than 742,000 imaging exams. The researchers compared median turnaround times before and after deployment.
The largest gain came for endotracheal tube malposition, where turnaround time dropped by about 76%, followed by rib fracture (53%) and brain vessel occlusion (29%). Pulmonary embolism and intracranial hemorrhage, however, saw no improvement. The researchers say this likely reflects higher proportions of follow-up or inpatient studies with known findings.
“As imaging volumes increase, timely identification of clinically significant findings is becoming more challenging,” the authors wrote. “Our experience shows that system-wide implementation of AI triage tools can support radiologists by prioritizing critical findings and improving workflow efficiency in high-volume settings. Since performance varied across applications, realizing these benefits requires targeted deployment and ongoing optimization as clinical workflows evolve.”
You’ll want to prioritize this Thursday morning session on your RSNA conference schedule.
