Nora Dajani MD, Presenter: Nothing to Disclose

Joel Garland Fletcher MD, Abstract Co-Author: Research grant, Siemens AG Grant, E-Z-EM, Inc License agreement, General Electric Company

Eric E. Williamson MD, Abstract Co-Author: Consultant, General Electric Company

Rathan Markandan Subramaniam MD, Abstract Co-Author: Nothing to Disclose

Cynthia H. McCollough PhD, Abstract Co-Author: Research grant, Siemens AG Research grant, RTI Electronics AB

Luca Bogoni PhD, Abstract Co-Author: Employee, Siemens AG, Malvern, PA

Garrett Spencer, Abstract Co-Author: Nothing to Disclose

Marcos Salganicoff PhD, Abstract Co-Author: Employee, Siemens AG

Osama I. Saba, Abstract Co-Author: Employee, Siemens AG

Theresa Nielson, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

To assess the performance of a proprietary computer-aided detection (CAD) system for the identification of acute pulmonary embolism (PE) using contrast-enhanced CT datasets obtained using a clinical PE protocol.

84 cases of acute PE were selected using a word text search of clinical reports. Chronic PE and non-PE protocol chest CT’s were excluded. To mimic routine clinical practice, we did not exclude datasets of suboptimal quality. PE studies employed 1.5–2 mm slice thickness and variable reconstruction kernels. Staff radiologists marked the 3D location of all acute PE’s, which served as the reference standard. A prototype CAD system (Siemens Medical Solutions, Malvern, PA) identified potential PE's on each scan. We used customized software to simultaneously compare CAD detections with the reference standard. CAD detections thought to represent true PE, but not detected by the radiologists' prospective interpretation, were recorded and included in the reference standard.

Per patient sensitivity was 87%(73/84; 77–93% 95%CI). Reasons for false negative exams included suboptimal pulmonary artery opacification (n=4), large effusions/atelectasis (n=3), and other causes (n=4). The per PE sensitivity was 59%(210/357; 54–64% 95%CI). False positive rate was 3.4 detections/case, with partial volume averaging, lymphoid tissue, and pulmonary vein accounting for the majority of cases (with 1.2, 0.9, and 0.4 detections/case, respectively). Importantly, the CAD system identified 28 detections (0.33/case) thought to represent true pulmonary emboli but missed by initial radiologist assessment.

Evaluation of CAD detections following radiologist assessment often resulted in additional pulmonary emboli being detected, even allowing for the fact that this prototype CAD system is not optimized for our acquisition and reconstruction parameters. Reported performance of this CAD system would improve if we preselected for only high quality datasets.

Computer-aided detection of pulmonary emboli is a promising tool that will likely improve the detection of pulmonary emboli, particularly for trainees, non-experts and fatigued radiologists.

Dajani, N, Fletcher, J, Williamson, E, Subramaniam, R, McCollough, C, Bogoni, L, Spencer, G, Salganicoff, M, Saba, O, Nielson, T, et al, , et al, , Computer-aided Detection of Acute Pulmonary Embolism Using Contrast-enhanced Multislice CT.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5013078.html