Can AI Improve Triage of CT Pulmonary Angiography Exams for Acute PE?

Emerging research suggests that artificial intelligence (AI) can improve the prioritization and efficiency of interpreting computed tomography pulmonary angiography (CTPA) exams for cases of suspected pulmonary embolism (PE).

For the retrospective study, recently published in the American Journal of Roentgenology, researchers assessed the impact of an AI-based worklist reprioritization tool (Brief Case for PE, Aidoc) in their review of 2,501 CTPA exams for 2,197 patients (mean age of 57.4) with suspected PE. Comparing 1,166 pre-AI CTPA exams versus 1,335 post-AI exams, the study authors utilized electronic medical record (EMR) and dictation system time stamps to examine the AI tool’s effect on wait time (the time between exam completion and initiation of the radiologist report); read time (the time between report initiation and report availability; and report turnround time (the total of wait and read times).

According to the study, use of the AI tool resulted in reprioritization of 148 CTPA exams (12.7 percent). For CTPA exams that were positive for PE, the researchers found no difference in read times with the AI tool but did note significant reductions in mean wait time (21.4 vs. 33.4 minutes) and mean radiology report turnaround time (47.6 vs. 59.9 minutes).

“Delays in reporting urgent findings may be particularly likely to occur for examinations ordered with routine priority, which are typically inserted at the bottom of the worklist with the implicit assumption that the probability of a positive finding is less than in a patient whose examination was ordered stat or urgent,” wrote lead study author Kiran Batra, M.D., an associate professor of radiology at the University of Texas Southwestern Medical Center in Dallas, and colleagues.

“A method that identifies acute findings on routine examinations and reprioritizes these examinations to the top of the worklist thus has the potential to substantially impact radiologist workflow and yield more rapid reporting of urgent findings.”

(Editor’s note: For related content, see “AI/CT Combination for Assessing Pulmonary Embolism Severity Gets FDA Nod” and “Can a New AI Tool Improve Detection of Incidental Pulmonary Embolism on CT?”)

The study authors pointed out that the AI tool does provide an additional series of images for review, which could possibly lead to increased time for interpreting images. However, the researchers found no significant difference in read times with the AI tool, noting a mean of 26.3 minutes with AI versus a mean of 26.5 minutes with pre-AI assessment).

“The lack of a significant difference in read times supports the worklist reprioritization as the underlying mechanism by which implementation of the AI tool led to earlier reporting of PE-positive examinations,” emphasized Batra and colleagues.

Beyond the inherent limitations of a retrospective study conducted at one institution, the researchers noted they did not assess factors such as radiologist experience and examination location in relation to the impact of the AI worklist tool. The study authors also pointed out that the sensitivity rate for the AI tool was lower than the rate reported in previous studies of the AI worklist tool. A failure with automated image transfer to the AI tool resulted in the exclusion of 223 CTPA exams in the post-AI period, according to Batra and colleagues.

(Editor’s note: For additional CT-related research, visit https://www.diagnosticimaging.com/topic/ct .)