HealthDay News — An artificial intelligence (AI) algorithm can help identify incidental pulmonary embolism (iPE) on conventional contrast-enhanced chest computed tomography (CT) examinations, according to a study published online July 13 in the American Journal of Roentgenology.
Kiran Batra, M.D., from the University of Texas Southwestern Medical Center in Dallas, and colleagues examined the diagnostic performance of an AI algorithm for detection of iPE among 2,555 patients with 3,003 conventional contrast-enhanced chest CT examinations performed between September 2019 and February 2020. All examinations positive by AI or a natural language-processing algorithm applied to the clinical reports underwent a multireader adjudication process to establish a reference standard for iPE.
The researchers found that the frequency of iPE was 1.3 percent based on the adjudication process. AI detected four iPEs that clinical reports had missed, while clinical reports identified seven iPEs missed by AI. Compared with clinical reports, AI exhibited significantly lower specificity (92.7 versus 99.8 percent) and positive predictive value (86.8 versus 97.3 percent), but differences in sensitivity (82.5 versus 90.0 percent) and negative predictive value (99.8 and 99.9 percent) were not significant. Neither sensitivity nor specificity varied significantly in association with age, sex, examination location, or cancer-related clinical scenario for AI.
“Potential clinical applications of the AI tool include serving as a second reader to help detect additional iPEs missed by radiologists or as a worklist triage and prioritization tool to allow earlier iPE detection and intervention,” the authors write.