Thursday, December 3 | 1:30 p.m.-2:30 p.m. | R6-SSCH08-2 | Room E451A
In this scientific presentation, researchers will share their experience with utilizing AI software for detection and triage of pulmonary embolism (PE).
Presenter Shlomit Goldberg-Stein of Northwell Health, and colleagues evaluated how AI-based PE detection and triage affected lead times, read times, and turnaround times for CT pulmonary angiography (CTPA) exams. Their analysis spanned clinical settings, staffing models, and work shifts in a large integrated health system.
In what they believe is the largest study of its kind, the researchers compared more than 90,000 CTPA exams performed before and after implementation of a U.S. Food and Drug Administration-cleared algorithm that automatically moves AI-positive PE exams up the worklist.
The algorithm correctly reprioritized 84% of PE-positive exams, with a false-positive elevation rate under 1%. The biggest reductions in lead time, or how long exams waited before being opened, came in resident-involved workflows (about 41 minutes) and inpatient settings (about 25 minutes). Attending-only interpretations, however, showed no differential effect.
“AI-based triage with active worklist reordering reduced [lead time] and [turnaround time] for PE-positive CTPAs, but effects varied substantially by setting and staffing model, with the greatest benefits in inpatient and resident-involved workflows,” the authors wrote. “Our findings support active worklist reprioritization, rather than passive alerts, as the preferred integration strategy for AI-driven PE triage.”
Drop by this Thursday session for more details.
