Twenty three (12 women and 11 men, mean age: 47 years, range: 14-83) consecutive patients suspected of having PE and who were referred for multislice CT angiography were included in this study. Ten patients underwent 16-slice CT pulmonary angiography (LightSpeed16, GE Healthcare, Milwaukee, WI) with the scanning protocol as follows: 120 kVp, 200-250 mAs, a beam collimation of 16 ×0.625 mm, a gantry rotation time of 500 ms, reconstruction of the slice thickness 1.0-1.25 mm, a pitch of 0.2-0.3 with a reconstruction interval of 0.5 or 0.6 mm. Thirteen patients underwent 64-slice CT pulmonary angiography (Brilliance 64, Philips Medical Systems, Best, the Netherlands) with the scanning protocol as follows: 120 kVp, 200-250 mAs, a beam collimation of 64×0.625 mm, a gantry rotation time of 420 ms, reconstruction of the slice thickness 0.6-1.25 mm, a pitch of 0.2-0.3 with a reconstruction interval of 0.4 or 0.6 mm. A field of view of 300-350 mm, a matrix size of 512×512 and 180° linear interpolation algorithms were used to reconstruct the images, resulting in an in-plane resolution of 0.58 mm× 0.58 mm-0.68 mm×0.68 mm. All of the CTPA scans were performed with intravenously injecting 60 to 100 ml of non-ionic contrast medium (Omnipaque 350, Schering, Berlin, Germany) and this was administered at a flow rate of 5 ml/s. The scan was initiated using a bolus tracking technique with a threshold of 100 HU over baseline at the main pulmonary trunk.

All the datasets were burned onto CDs from the MSCT scanner and these were transferred to a separate workstation for generating the VIE images. The MSCT volume data was converted from the original DICOM (Digital Imaging and Communication in Medicine) images with using the commercially available software Analyze V 7.0 (Analyze V, AnalyzeDirect, Inc., Lexana, KS). The details of generating the VIE images of the artery lumen have been described elsewhere (14, 15). Similar to the previous studies, we used a CT number thresholding technique to produce endoscopic views of the pulmonary artery and its branches, as well as endoscopic views of the abnormal changes (14, 15). This was determined by selecting the appropriate threshold value through measuring the CT attenuation at the left and right main pulmonary arteries. An upper threshold of 200-250 HU was applied to remove the contrast-enhanced blood from the vessel while keeping the thrombus and vessel lumen.

To assess the image quality of the multislice CTPA images, the image noise and the CT attenuation of the main pulmonary artery, as well as the signal to noise ratio (SNR), were determined in each patient. Image noise was defined as the standard deviation (SD) of the CT density in a region of interest (ROI) that was placed in the pulmonary artery (Fig. 1). The ROI was chosen to be placed in the centre of the pulmonary artery with a minimum area of 50 mm² while carefully avoiding inclusion of the artery wall or thrombus to prevent any partial volume effects. The SNR was determined by dividing the mean attenuation by the image noise.

The mean SNR measured at the left and right pulmonary arteries was 14.2 ± 4.6 and 14.4 ± 5.4, respectively, for the 16-slice CTPA protocol, and it was 24.7 ± 10.7 and 20 ± 7.4, respectively, for the 64-slice CTPA protocol. The SNR was significantly different between the 16- and 64-slice CTPA protocols when it was measured at the left pulmonary artery (p = 0.014), but there was no significant difference when the SNR was measured at the right pulmonary artery (p = 0.134).

Virtual intravascular endoscopy was successfully generated in all of the patients. The VIE image quality was determined by the SNR measured at the pulmonary artery and its branches. The VIE image of the pulmonary artery with a high SNR shows a smooth intraluminal appearance (Fig. 2), while the VIE image of the pulmonary artery with a low SNR shows a smooth irregular appearance of the pulmonary artery wall (Fig. 3).

In this study, pulmonary embolism is recognized as a thrombus within the pulmonary vessel and it has CT attenuation much lower (less than 50 HU) than that of the contrast enhanced blood in the pulmonary artery (300-400 HU). Thus, when applying a CT threshold (the upper threshold was normally chosen to be 200-300 HU) to remove the contrast-enhanced blood from the pulmonary artery, the low attenuation thrombus was kept together with the pulmonary wall and lumen in the generated VIE views. Thus, pulmonary embolism appears as an intraluminal protrusion arising from the wall (Fig. 4).

Although our study is based on a small number of cases, our findings provide insight into the intraluminal appearances of pulmonary embolism with the aid of VIE visualization. In contrast to the conventional 2D or 3D extraluminal views, VIE allows for uniquely demonstrating the thrombus, whether it is located in the main pulmonary artery or in the lobar or segmental branches. This is especially useful for identifying a thrombus in the segmental or subsegmental arteries as an embolism located at these side branches is sometimes difficult to detect on CT images. Combined with the 2D views, VIE is able to confirm the thrombus in the distal pulmonary branches. Moreover, VIE visualization allows for a virtual fly-through of the entire pulmonary tree, which is valuable for assessing the extent of disease. However, the VIE images must be combined with the corresponding orthogonal views to confirm the segmental or subsegmental pulmonary embolism as it is difficult to determine the exact location of these branches based on the VIE image alone.

Virtual intravascular endoscopy has been reported to be useful to follow-up the endovascular stent graft repair of aortic aneurysm (14-18). VIE has additional value to demonstrate the intraluminal appearance of aortic stent grafts or fenestrated vessel stents in relation to the aortic branches, and mainly the renal arteries. VIE can enhance endovascular specialists' understanding of the treatment outcome of an endovascular repair of aortic aneurysm. To the best of our knowledge, there have been no studies performed regarding the application or diagnostic value of VIE for assessing pulmonary embolism. Our preliminary experience shows that VIE is not only able to demonstrate the intraluminal appearance of a pulmonary embolism, but it also assesses the extent of disease. The main potentials of providing direct demonstration of the intraluminal thrombus or the arterial wall changes by VIE lie in the following aspects, which we believe are useful from the clinical perspective:

• Accurately determining the location and extent of the pulmonary embolism, and especially for the cases with segmental or subsegmental embolism, where conventional 2D or 3D visualizations have limitations.

• Diagnosing and confirming a pulmonary embolism in indeterminate cases.

• Quantitatively evaluating the pre- and post-treatment of a thrombus and so this contributes to assessing the treatment outcome and predicting the progress of disease.

• Dynamically assessing a pulmonary embolism via a virtual fly-through instead of using the static views, and this is considered to be valuable for assessing the severely occluded or near totally occluded branches.

The VIE image quality is determined by the degree of contrast enhancement in the pulmonary artery, yet the visualization of the right pulmonary artery is affected, to some extent, by the high-density contrast medium in the superior vena cava. Artifacts, and especially those at the right pulmonary artery, may interfere with the VIE visualization of the intraluminal appearance of a pulmonary embolism. Thus, artifacts could cause difficulty to differentiate the thrombus from the artifacts. However, a pulmonary embolism normally presents as a protruding sign from the intact artery lumen without any disruption of the artery wall. In contrast, the artifacts result in an irregular appearance in the artery lumen or they may even lead to discontinuity of the lumen. Streaking artifact can be distinguished from a pulmonary embolism by recognizing its non-anatomic, poorly defined, radiating nature (Fig. 3A). The use of a 'saline chase' in CTPA could be an effective strategy to dilute the contrast density in the superior vena cava while still acquiring the diagnostic images, but it minimizes the degree of artifacts.

Researchers have previously reported that respiratory motion artifacts can cause the misdiagnosis of pulmonary embolism, and respiratory motion artifacts are the most common cause of an indeterminate CTPA (19). Jones and Wittram (20) found motion artifacts to be responsible for 74% of the indeterminate CTPAs. Since the main cause of motion during CTPA is the inability of dyspneic patients to hold their breath during the scanning time, the development and increased use of 16- and 64-slice CT with their substantially faster scanning times should lower the number of indeterminate CTPA cases. Yavas et al. (21) in their recent study reported that the thoracic CT images were nondiagnostic for two of 112 patients due to lung motion. No motion artifacts were present in our patients, and this was mainly due to the fast scanning technique available with using 16- and 64-slice CT in this study.

In conclusion, we believe that VIE could be used as a complementary tool together with the conventional 2D visualizations for evaluating pulmonary embolisms, in terms of the quantitative assessment of thrombus changes after medical treatment, the intraluminal changes of the arterial wall that are due to thrombus and the extent of a pulmonary embolism.