Planar pulmonary scintigraphy is still regularly performed for the evaluation of pulmonary embolism (PE). However, only about 50-80% of cases can be resolved by this approach. This study evaluates the ability of tomographic acquisition (single photon emission computed tomography, SPECT) of the perfusion scan to improve the radionuclide diagnosis of PE. One hundred and fourteen consecutive patients with a suspicion of PE underwent planar and SPECT lung perfusion scans as well as planar ventilation scans. The final diagnosis was obtained by using an algorithm, including D-dimer measurement, leg ultrasonography, a V/Q scan and chest spiral computed tomography, as well as the patient outcome. A planar perfusion scan was considered positive for PE in the presence of one or more wedge shaped defect, while SPECT was considered positive with one or more wedge shaped defect with sharp borders, three-plane visualization, whatever the photopenia. A definite diagnosis was achieved in 70 patients. After exclusion of four 'non-diagnostic' SPECT images, the prevalence of PE was 23% (n =15). Intraobserver and interobserver reproducibilities were 91%/94% and 79%/88% for planar/SPECT images, respectively. The sensitivities for PE diagnosis were similar for planar and SPECT perfusion scans (80%), whereas SPECT had a higher specificity (96% vs 78%; P =0.01). SPECT correctly classified 8/9 intermediate and 31/32 low probability V/Q scans as negative. It is concluded that lung perfusion SPECT is readily performed and reproducible. A negative study eliminates the need for a combined V/Q study and most of the 'non-diagnostic' V/Q probabilities can be solved with a perfusion image obtained by using tomography.