The effect of vascular curvature on three-dimensional reconstruction of intravascular ultrasound images.

Abstract

To characterize the effect of vessel curvature on the geometric accuracy of conventional three-dimensional reconstruction (3DR) algorithms for intravascular ultrasound image data. A common method of 3DR for intravascular ultrasound image data involves geometric reassembly and volumetric interpolation of a spatially related sequence of tomographic cross sections generated by an ultrasound catheter withdrawn at a constant rate through a vascular segment of interest. The resulting 3DR is displayed as a straight segment, with inherent vascular curvature neglected. Most vascular structures, however, are not straight but curved to some degree. For this reason, vascular curvature may influence the accuracy of computer-generated 3DR. We collected image data using three different intravascular ultrasound catheters (2.9 Fr, 4.3 Fr, 8.0 Fr) during a constant-rate pullback of 1 mm/sec through tubing of known diameter with imposed radii of curvature ranging from 2 to 10 cm. Image data were also collected from straight tubing. Image data were digitized at 1.0-mm intervals through a length of 25 mm. Two passes through each radius of curvature were performed with each intravascular ultrasound catheter. 3DR lumen volume for each radius of curvature was compared to that theoretically expected from a straight cylindrical segment. Differences between 3DR lumen volume of theoretical versus curved (actual) tubes were quantified as absolute percentage error and categorized as a function of curvature. Tubing deformation error was quantified by quantitative coronary angiography (QCA). Volumetric errors ranged from 1% to 35%, with an inverse relationship demonstrated between 3DR lumen volume and segmental radius of curvature. Higher curvatures (r < 6.0 cm) induced greater lumen volume error when compared to lower curvatures (r > 6.0 cm). This trend was exhibited for all three catheters and was shown to be independent of tubing deformation artifacts. QCA-determined percentage diameter stenosis indicated no deformation error as a function of curvature. Total volumetric error contributed by tubing deformation was estimated to be 0.05%. Catheter-dependent geometrical error arises in three-dimensionally reconstructed timed linear pullbacks of intravascular ultrasound images due in part to uniplanar vascular curvature. Three-dimensional reconstruction of timed linear pullbacks is robust for vessels with low radii of curvature; however, careful interpretation of three-dimensional reconstructions from timed linear pullbacks for higher radii of curvature is warranted. These data suggest that methods of spatially correct three-dimensional reconstruction of intravascular ultrasound images should be considered when more pronounced vascular curvature is present.