Purpose: To determine diagnostic accuracy of four-channel multi-detector row computed tomography (CT) in emergency room and inpatient populations suspected of having acute pulmonary embolism (PE) who prospectively underwent both CT and pulmonary arteriography (PA).
Materials and methods: Patients referred for PA to assess suspected PE were eligible. Institutional review board approval and written informed consent were obtained. All patients underwent CT and PA within a 48-hour period. For CT, 4 x 2.5-mm collimation was used. Three readers independently evaluated each study for PE presence. PE status, vessel level, and lobar location were determined by means of majority rule, and interobserver agreement (kappa) was calculated for PE status, as assessed with each modality. Sensitivity and specificity of CT were calculated by using PA as the reference standard. Two radiologists later reviewed false-positive CT studies.
Results: The study group comprised 93 patients (median age, 56 years; range, 19-88 years). Sensitivity, specificity, and accuracy of CT were 100%, 89%, and 91%, respectively. kappa values were 0.71 and 0.83 for CT and PA, respectively, and were not significantly different between modalities. At PA, 18 patients (19%) had PE at 50 vessel levels (five main and/or interlobar, 24 segmental, and 21 subsegmental), 17 (94%) of which had PE at multiple sites. At CT, 26 patients (28%) had PE at 71 vessel levels (24 main and/or interlobar, 33 segmental, and 14 subsegmental). Twenty patients (77%) had PE at multiple sites. Review of eight false-positive CT studies showed an appearance highly suggestive of acute PE in three patients, chronic PE in one, and no PE in three; one study was inconclusive. CT better demonstrated large-level vessel involvement (P < .01), while PA better demonstrated small-level vessel involvement (P < .01).
Conclusion: Multi-detector row CT has an accuracy of 91% in the depiction of suspected acute PE when conventional PA is used as the reference standard.
(c) RSNA, 2004.