Computed tomography diagnosis of pulmonary infarction in acute pulmonary embolism

Introduction: Pulmonary infarction is a common sequela of pulmonary embolism (PE) but lacks a diagnostic reference standard. CTPA in the setting of acute PE does not reliably differentiate infarction from other consolidations, such as reversible alveolar hemorrhage or atelectasis. We aimed to assess the diagnostic accuracy for recognizing pulmonary infarction on CT in the acute phase of PE, with follow-up CT as reference. Materials and methods: Initial and follow-up CT scans of 33 patients with acute PE were retrospectively assessed. Two radiologists independently evaluated the presence and size of suspected pulmonary infarction on the initial CT. Confirmation of infarction was established by detection of residual densities on follow-up CT. Sensitivity, specificity and interobserver variability were calculated. Results: In total, 60 presumed infarctions were found in 32 patients, of which 34 infarctions in 21 patients could be confirmed at follow-up. On patient-level, observers' sensitivity/specificity were 91 %/9 %, and 73 %/46 %, respectively, with interobserver agreement by Kappa's coefficient of 0.17. Confirmed infarctions were usually larger than false positive lesions (median approximate volume of 6.6 mL [IQR 0.84-21.3] vs. 1.3 mL [IQR 0.57-6.5], p = 0.040), but still small. An occluding thrombus in a supplying vessel was predictive for confirmed infarction (OR 11, 95%CI 2.1-55), but was not discriminative. Conclusions: Pulmonary infarction is a common finding in acute PE, and generally small. Radiological identification of infarction was challenging, with considerable interobserver variability. Complete obstruction of the supplying (sub)segmental pulmonary artery was found as the strongest predictor for pulmonary infarction but was not demonstrated to be discriminative.