The purpose of our animal study was to evaluate a new computed tomography (CT) subtraction technique for visualization of perfusion defects within the lung parenchyma in subsegmental pulmonary embolism (PE). Seven healthy pigs were entered into a prospective trial. Acute PE was artificially induced by fresh clot material prior to the CT scans. Within a single breath-hold, whole thorax CT scans were performed with a 16-slice multidetector-row CT scanner (SOMATOM Sensation 16; Siemens, Forchheim, Germany) before and after intravenous application of 80 ml of contrast medium with a flow rate of 4 ml/s, followed by a saline chaser. The scan parameters were 120 kV and 100 mAs(eff), using a thin collimation of 16x0.75 mm and a table speed/rotation of 15-18 mm (pitch, 1.25-1.5; rotation time, 0.5 s). Axial source images were reconstructed with an effective slice thickness of 1 mm (overlap, 30%). A new automatic subtraction technique was used. After 3D segmentation of the lungs in the plain and contrast-enhanced series, threshold-based extraction of major airways and vascular structures in the contrast images was performed. This segmentation was repeated in the plain CT images segmenting the same number of vessels and airways as in the contrast images. Both scans were registered onto each other using nonrigid registration. After registration both image sets were filtered in a nonlinear fashion excluding segmented airways and vessels. After subtracting the plain CT data from the contrast data the resulting enhancement images were color-encoded and overlaid onto the contrast-enhanced CT angiography (CTA) images. This color-encoded combined display of parenchymal enhancement of the lungs was evaluated interactively on a workstation (Leonardo, Siemens) in axial, coronal and sagittal plane orientations. Axial contrast-enhanced CTA images were rated first, followed by an analysis of the combination images. Finally, CTA images were reread focusing on areas with perfusion deficits indicating PE on the color-coded enhancement display. Subtraction was feasible for all seven studies. In one animal, opacification of the pulmonary arteries was suboptimal owing to heart insufficiency. In the remaining six pigs, a total of 37 perfusion defects were clearly assessable downstream of occluded subsegmental arteries, showing lower or missing enhancement compared with normally perfused lung parenchyma. Indeterminate findings from CTA showed typical PE perfusion defects in four out of six cases on CT subtraction. Additionally, 22 peripheral triangular-shaped enhancement defects were delineated. Nine of these findings were reclassified as definitely being caused by PE on second reading of the CTA data sets. Our initial results have shown that this new subtraction technique for perfusion imaging of PE is feasible, using routine contrast delivery. Dedicated examination protocols are mandatory for adequate opacification of the pulmonary arteries and for optimization of data sets for subsequent subtraction. Perfusion imaging allows a comprehensive assessment of morphology and function, providing more accurate information on acute PE.