Deep learning-based automated detection of pulmonary embolism on CT pulmonary angiograms: No significant effects on report communication times and patient turnaround in the emergency department nine months after technical implementation

• DOI: 10.1016/j.ejrad.2021.109816
• Publication Date: 2021-06-11T08:41:44Z
• Authors: Lena Schmuelling; Fabian C. Franzeck; Christian H. Nickel; Gregory Mansella; Roland Bingisser; Noemi Schmidt; Bram Stieltjes; Jens Bremerich; Alexander W. Sauter; Thomas Weikert; Gregor Sommer
• Keywords: Turnaround time

ObjectivesRapid communication of CT exams positive for pulmonary embolism (PE) is crucial timely initiation anticoagulation and patient outcome. It unknown if deep learning automated detection PE on Pulmonary Angiograms (CTPA) in combination with worklist prioritization an electronic notification system (ENS) can improve times turnaround the Emergency Department (ED).MethodsIn 01/2019, ENS allowing direct between radiology ED was installed. Starting 10/2019, CTPAs were processed by a (DL)-powered algorithm PE. acquired 04/2018 06/2020 (n = 1808) analysed. To assess impact DL-algorithm, report reading (RRT), time (RCT), to (TTA), (TAT) compared three consecutive periods. Performance measures calculated per exam level (sensitivity, specificity, PPV, NPV, F1-score), written reports review as ground truth.ResultsSensitivity 79.6 % (95 %CI:70.8−87.2%), specificity 95.0 %CI:92.0−97.1%), PPV 82.2 %CI:73.9−88.3), NPV 94.1 %CI:91.4–96 %). There no statistically significant reduction any observed (RRT, RCT, TTA, TAT).ConclusionDL-assisted ENS-assisted results referring physicians technically work. However, mere clinical introduction these tools, even they exhibit good performance, not sufficient achieve effects performance measures.