New quantification methods for carotid intraplaque neovascularization in contrast enhanced ultrasound

摘要

Carotid intraplaque neovascularization (IPN) has been associated with progressive atherosclerotic disease and plaque vulnerability. Therefore, its accurate quantification might allow early detection of plaque vulnerability. Contrast enhanced ultrasound (CEUS) can detect these small microvessels. To quantify IPN, we developed quantitative methods based on time intensity curve (TIC) and maximum intensity projection (MIP), micro-vascular structure analysis (VSA), and statistical segmentation (SS). Plaque region of interest (ROI) is manually drawn and motion compensation is applied before each analysis. In TIC and MIP, we examine perfusion dynamics and regions within plaques. In VSA, we detect and track contrast spots to examine the microvessel network. In SS, we classify plaque intensities into different components for quantification of IPN. Through an iterative expectation-maximization algorithm, plaque pixels are initially labeled into artifacts, contrast, intermediate, and background class. Next, spatiotemporal and neighborhood information is used to relabel intermediate class pixels, remove artifacts and correct false-contrast. From the applied analyses, we derived several parameters - e.g. MIP based IPN surface area (MIPNSA), MIP based surface ratio (MIPNSR), SS based IPN surface area (SSIPNSA), plaque mean intensity, mean plaque contrast percentage, and number of microvessels (MVN) - and compared them to consensus of visual grading of IPN by two independent physicians. We analyzed 45 carotid arteries with stenosis. To verify if SSIPNSA improves the suppression of artifacts, we analyzed 8 plaques twice, with saturation artifacts included and excluded from the ROI. Five parameters were found to be significantly correlated to visual scoring and may thus have the potential to replace qualitative visual scoring and to measure the degree of carotid IPN. The MIPNSA & SSIPNSA parameters gave the best distinction between visual scores. SSIPNSA proved less sensitive - or artifacts than MIPNSA.