Pulmonary embolism: optimization of small pulmonary artery visualization at multi-detector row CT.

Abstract

To compare the frequency of well-visualized pulmonary arteries according to anatomic level by using different collimation with single- and multi-detector row computed tomography (CT) in patients suspected of having acute pulmonary embolism. Sixty patients were examined with one of three techniques (20 patients each). Group 1 was examined with single-detector row CT with 3-mm collimation and 1.3-1.6 pitch; groups 2 and 3, with multi-detector row CT with 2.5- and 1.25-mm collimation, respectively. Three thoracic radiologists independently reviewed examination findings to determine if each main, lobar, segmental, and subsegmental artery was well visualized for presence of pulmonary embolism. chi2 tests were performed. For well-visualized vessels, the presence and/or absence of pulmonary embolism was recorded and kappa statistic was determined. Reader 1 scored 95% (114 of 120), 96% (115 of 120), and 99% (119 of 120) of lobar arteries (P >.05); 76% (304 of 400), 86% (346 of 400), and 91% (363 of 400) of segmental arteries (P <.001); and 37% (300 of 800), 56% (448 of 800), and 76% (608 of 800) of subsegmental arteries as well visualized (P <.001) using techniques 1, 2, and 3, respectively. Reader 2 scored 97% (116 of 120), 95% (114 of 120), and 99% (119 of 120) of lobar arteries (P >.05); 77% (308 of 400), 87% (349 of 400), and 93% (371 of 400) of segmental arteries (P <.001); and 39% (310 of 800), 53% (422 of 800), and 78% (621 of 800) of subsegmental arteries (P <.001) as well visualized using techniques 1, 2, and 3, respectively. Reader 3 scored 86% (103 of 120), 82% (98 of 120), and 91% (109 of 120) of lobar arteries (P >.05); 63% (252 of 400), 70% (280 of 400), and 85% (339 of 400) of segmental arteries (P <.001); and 39% (310 of 800), 56% (451 of 800), and 71% (572 of 800) of subsegmental arteries (P <.001) as well visualized using techniques 1, 2, and 3, respectively. Sixteen patients had pulmonary embolism. Interobserver agreement for detection of pulmonary embolism was significantly better for segmental and subsegmental arteries for all readers with technique 3 (segmental, kappa = 0.79-0.80; subsegmental, kappa = 0.71-0.76) than that with technique 1 (segmental, kappa = 0.47-0.75; subsegmental, kappa = 0.28-0.54). Multi-detector row CT at 1.25-mm collimation significantly improves visualization of segmental and subsegmental arteries and interobserver agreement in detection of pulmonary embolism.