OBJECTIVE To evaluate the accuracy of spiral computed tomography (CT) in the noninvasive diagnosis of pulmonary embolism (PE). METHODS A prospective study was performed in 75 patients who were evaluated with spiral CT and pulmonary angiography of each lung to detect central PE; 25 of the patients also underwent ventilation-perfusion (V-P) scanning. RESULTS Spiral CT scans were technically suboptimal in three patients. CT and angiographic findings were negative for PE in 25 patients; one patient had false-negative CT findings. Findings from both studies were positive in 39 patients. CT findings of 188 central emboli corresponded exactly to those of angiography. Ten emboli were depicted only on CT scans, whereas seven emboli were identified only on angiograms because of inadequate depiction of the pulmonary arteries in the plane of the CT scans (n = 5) or because of misinterpretation of CT findings (n = 2). The prospective sensitivity of CT was 91%, the specificity was 78%, the positive predictive value was 100%, and the negative predictive value was 89%. Technical failures (n = 3) and inconclusive CT findings (n = 7) were the major limitations of spiral CT. Spiral CT enabled accurate classification of PE in 16 patients with indeterminate (n = 7) and low (n = 9) probability of PE on V-P scans. CT demonstrated central PE in two patients with normal V-P scans. CONCLUSIONS Spiral CT can reliably depict central PE and may be introduced into the classic diagnostic algorithms.