Unsuspected pulmonary embolism: prospective detection on routine helical CT scans.

PURPOSE: To determine the prevalence of unsuspected pulmonary embolism (PE) on routine thoracic helical computed tomographic (CT) scans and to quantify the improvement in PE detection by using a cine-paging mode on a workstation instead of hard-copy review. MATERIALS AND METHODS: Seven hundred eighty-five patients referred for routine contrast medium-enhanced thoracic CT within 9 months were prospectively recruited. Helical CT was performed. Studies were prospectively interpreted by four radiologists. Two radiologists performed routine, undirected, hard-copy consensus review for official interpretation; two of three thoracic radiologists independently performed a dedicated workstation-based search for PE. The presence of PE involving the main, lobar, or segmental pulmonary arteries was assigned a score of 1-5 (1 = definitely negative, 5 = definitely positive) by each independent reviewer. Patients with a score of 4 or 5 underwent lower-extremity ultrasound, ventilation-perfusion scintigraphy, or both, followed by pulmonary CT angiography if the findings were still equivocal. RESULTS: Twelve (1.5%) of the 785 patients had unsuspected PE, with an inpatient prevalence of 5% (eight of 160) and an outpatient prevalence of 0.6% (four of 625). Of the 12 patients with unsuspected PE, 10 (83%) had cancer. Of the 81 inpatients with cancer, seven (9%) had unsuspected PE. A dedicated workstation-based search resulted in detection of PE in three more patients (25%) than did hard-copy interpretation. CONCLUSION: The prevalence of unsuspected PE was highest among inpatients with cancer. A directed, workstation-based search can improve the PE detection rate over that with hard-copy review.