Validation of Automated Border Detection in Intravascular Ultrasound Images

Intravascular ultrasound (IVUS) imaging provides cross-sectional views of the vessel lumen; however, lumen measurements still rely on operator-dependent border delineation and time-consuming lumen tracings. We tested a new system for automated lumen border detection in IVUS images based on acoustic quantification of blood and vessel wall. In 10 rabbits, 29 segments of the aorta were imaged in vivo using a 2.9-Fr IVUS catheter. IVUS images were obtained during motorized pullbacks of aortic segments of 18 mm length. Automated measurements of lumen dimensions were compared to automated measurements of a second pullback through the same segment, lumen measurements derived from visual border tracings in IVUS images, and to quantitative angiography. The automated system showed good reproducibility: Correlations for repeated measurements of lumen area, maximal and minimal lumen diameters were r = 0.97, r = 0.91, and r = 0.93, respectively. Automated measurements also correlated well to visual image analysis (lumen area, r = 0.97; maximal lumen diameter, r = 0.89; minimal lumen diameter, r = 0.89) and to angiographic measurements (lumen area, r = 0.93; lumen diameter, r = 0.95). In 12% of the images, the automated system overestimated lumen dimensions because of weak wall signals in the presence of echolucent structures next to the wall. Signal artifacts from the IVUS catheter itself or strong blood backscatter resulted in lumen underestimation in 6% of the images. Over- and underestimation of lumen by the border detection system were often associated with eccentric catheter position. Thus, lumen measurements in vivo IVUS images can be performed using an automated border detection system based on acoustic quantification of blood and vessel wall. The system allows reproducible and accurate measurements of lumen area and diameters. (ECHOCARDIOGRAPHY, Volume 13, November 1996)