2020/08/10 by Richard D. White, Barbaros S. Erdal, White, Richard D. +21
Engineering · Medicine · #Advanced X-ray and CT Imaging #Cardiac Imaging and Diagnostics #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #FOS: Physical sciences #I.2.10 #I.5.2 #I.5.4 #Image and Video Processing (eess.IV) #Medical Physics (physics.med-ph) #Radiation Dose and Imaging #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2008.04802
openalex publication_date 2020/08/10 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Coronary Computed Tomography Angiography (CCTA) evaluation of chest-pain\npatients in an Emergency Department (ED) is considered appropriate. While a\nnegative CCTA interpretation supports direct patient discharge from an ED,\nlabor-intensive analyses are required, with accuracy in jeopardy from\ndistractions. We describe the development of an Artificial Intelligence (AI)\nalgorithm and workflow for assisting interpreting physicians in CCTA screening\nfor the absence of coronary atherosclerosis. The two-phase approach consisted\nof (1) Phase 1 - focused on the development and preliminary testing of an\nalgorithm for vessel-centerline extraction classification in a balanced study\npopulation (n = 500 with 50% disease prevalence) derived by retrospective\nrandom case selection; and (2) Phase 2 - concerned with simulated-clinical\nTrialing of the developed algorithm on a per-case basis in a more real-world\nstudy population (n = 100 with 28% disease prevalence) from an ED chest-pain\nseries. This allowed pre-deployment evaluation of the AI-based CCTA screening\napplication which provides a vessel-by-vessel graphic display of algorithm\ninference results integrated into a clinically capable viewer. Algorithm\nperformance evaluation used Area Under the Receiver-Operating-Characteristic\nCurve (AUC-ROC); confusion matrices reflected ground-truth vs AI\ndeterminations. The vessel-based algorithm demonstrated strong performance with\nAUC-ROC = 0.96. In both Phase 1 and Phase 2, independent of disease prevalence\ndifferences, negative predictive values at the case level were very high at\n95%. The rate of completion of the algorithm workflow process (96% with\ninference results in 55-80 seconds) in Phase 2 depended on adequate image\nquality. There is potential for this AI application to assist in CCTA\ninterpretation to help extricate atherosclerosis from chest-pain presentations.\n