The objective of this prospective study was to evaluate the sensitivity, specificity, positive and negative predictive values, and interobserver agreement in the diagnosis of pulmonary embolism with helical CT, compared with pulmonary angiography, for both global results and for selective vascular territories. Helical CT and pulmonary angiography were performed on 66 consecutive patients with clinical suspicion of pulmonary embolism. The exams were blindly interpreted by a vascular radiologist and by two independent thoracic radiologists. Results were analyzed for the final diagnosis as well as separately for 20 different arterial territories in each patient. Pulmonary angiography revealed embolism in 25 patients (38%); 48% were main, 28% lobar, 16% segmental, and 8% subsegmental. The sensitivity, specificity, and positive and negative predictive values of helical CT for observer 1 were, respectively, 91, 81.5, 75, and 94%; in 7.5% of the patients the exam was considered indeterminate. For observer 2 the values were, respectively, 88, 86, 81.5, and 91%; in 9% of the patients the exam was considered indeterminate. Main arteries were considered as non-valuable in 0-0.8%, the lobar in 1.5%, the segmental in 7.5-8.5%, and the subsegmental in 55-60%. Interobserver agreement for the final diagnosis was 80% (kappa 0.65). For each vascular territory, this was 98% (kappa 0.91) for main arteries, 92% (kappa 0.78) for lobar arteries, 79% (kappa 0.56) for segmental arteries, and 59% (kappa 0.21) for subsegmental arteries. Helical CT is a reliable method for pulmonary embolism diagnosis, with good interobserver agreement for main, lobar, and segmental territories. Worse results are found for subsegmental arteries, with high incidence of non-valuable branches and poor interobserver agreement.