Intravascular ultrasound (IVUS) imaging is a clinical tool that permits direct visualization of vascular pathology. It has been increasingly used to evaluate lumen and plaque morphology in coronary arteries. Conventional manual evaluation is tedious and time-consuming.;We have developed an automated approach to segmentation of arterial lumen and plaque in two-dimensional images and three-dimensional pullback sequences. The method incorporates knowledge from vascular anatomy, ultrasound physics, and image processing and uses an optimal graph-searching border detection approach. We have also developed a new method for automated determination of plaque composition. The method uses statistical pattern recognition approach to plaque characterization.;To validate our methods, IVUS images and image sequences were acquired from coronary arteries in vivo and in vitro. To assess performance of automated segmentation, computer-detected borders were compared to observer-defined borders. Quantitative measurements were derived to evaluate the method. High correlations were found between computer-detected and observer-defined lumen and plaque area in 38 individual images (r = 0.98, y = 0.98x + 0.04, r = 0.98, y = 1.00x + 0.36), and between original lumen and plaque area in 20 ECG-gated pullback sequences (r = 0.98, y = 1.01x + 1.51, r = 0.94, y = ;Our method clearly demonstrates the feasibility of automated segmentation and tissue characterization in 2D and 3D IVUS images.
展开▼
