Accuracy of a deep neural network for automated pulmonary embolism detection on dedicated CT pulmonary angiograms.
Authors
Affiliations (22)
Affiliations (22)
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Radiology Department, Medical Imaging Centre, Semmelweis University, Korányi Sándor utca 2, 1083 Budapest, Hungary. Electronic address: [email protected].
- Siemens Healthineers, 755 College Rd E, Princeton, NJ 08540, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Experimental Imaging Center, IRCCS San Raffaele Scientific Institute, Via Olgettina, 60, 20132 Milan, Italy. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Heart and Vascular Center, Semmelweis University, Gaál József út 9, 1122 Budapest, Hungary. Electronic address: [email protected].
- Division of Cardiology, Department of Medicine, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Department of Diagnostic and Interventional Radiology, Medical Center, Faculty of Medicine, University of Freiburg, Hugstetter Straße 55, Freiburg im Breisgau 79106, Germany. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Department of Diagnostic and Interventional Radiology, University Hospital Bonn, Venusberg-Campus 1, 53127 Bonn, Germany. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Department of Medical Surgical Sciences and Translational Medicine, Sapienza University of Rome - Radiology Unit - Sant'Andrea University Hospital, Via di Grottarossa, 1035-1039, 00189 Rome, Italy. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Department of Radiology, University Hospital, LMU Munich, Marchioninistraße 15, 81377 Munich, Germany. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
- Siemens Healthineers, Nine, Bulevardul Gării 13A, Brașov 500227, Romania. Electronic address: [email protected].
- Siemens Healthineers, 755 College Rd E, Princeton, NJ 08540, USA. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Siemens Medical Solutions, 40 Liberty Blvd, Malvern, PA 19355, USA. Electronic address: [email protected].
- Radiology Department, Medical Imaging Centre, Semmelweis University, Korányi Sándor utca 2, 1083 Budapest, Hungary. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA; Department of Diagnostic and Interventional Radiology, University Medical Center of the Johannes Gutenberg-University, Langenbeckstraße 1, Mainz 55131, Germany; German Centre for Cardiovascular Research, Mainz, Germany. Electronic address: [email protected].
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425, USA. Electronic address: [email protected].
Abstract
To assess the performance of a Deep Neural Network (DNN)-based prototype algorithm for automated PE detection on CTPA scans. Patients who had previously undergone CTPA with three different systems (SOMATOM Force, go.Top, and Definition AS; Siemens Healthineers, Forchheim, Germany) because of suspected PE from September 2022 to January 2023 were retrospectively enrolled in this study (n = 1,000, 58.8 % women). For detailed evaluation, all PE were divided into three location-based subgroups: central arteries, lobar branches, and peripheral regions. Clinical reports served as ground truth. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were determined to evaluate the performance of DNN-based PE detection. Cases were excluded due to incomplete data (n = 32), inconclusive report (n = 17), insufficient contrast detected in the pulmonary trunk (n = 40), or failure of the preprocessing algorithms (n = 8). Therefore, the final cohort included 903 cases with a PE prevalence of 12 % (n = 110). The model achieved a sensitivity, specificity, PPV, and NPV of 84.6, 95.1, 70.5, and 97.8 %, respectively, and delivered an overall accuracy of 93.8 %. Among the false positive cases (n = 39), common sources of error included lung masses, pneumonia, and contrast flow artifacts. Common sources of false negatives (n = 17) included chronic and subsegmental PEs. The proposed DNN-based algorithm provides excellent performance for the detection of PE, suggesting its potential utility to support radiologists in clinical reading and exam prioritization.