Dual Fuzzy-Snake model for IMT measurement in carotid ultrasound images of low- and high-risk patients

Abstract

The carotid artery intima-media thickness (cIMT) is commonly deemed an indicator for cardiovascular diseases. Several methods were proposed for the automatic estimation of the IMT on ultrasound carotid images, in order to obtain accurate, objective and repetitive values. One of the most used techniques for image segmentation is the snake, often requiring a preliminary manual phase of parameter tuning and profile initialization. In this study we present a new segmentation method for automatic detection of lumen-intima (LI) and media-adventitia (MA) interfaces. This technique is based on the simultaneously evolution of the two snakes under the action of a Fuzzy Inference System (FIS). We tested our method for images both with normal carotid wall and with plaques. The performances of the Fuzzy-Snake system result better than the classical snake implementation, showing no trends between errors and IMT values. In plaqued vessels, the FIS driven snake improves the detection of the LI compared to classical snakes.