Computer-aided quantification of contrast agent spatial distribution within atherosclerotic plaque in contrast-enhanced ultrasound image sequences

Abstract

Recent studies have revealed that the contrast-enhanced ultrasound (CEUS) correlates to the presence and degree of intraplaque neovascularization as determined histologically. However, most studies used a qualitative system to visually grade CEUS images. A computer-aided method is proposed for objective and convenient quantification of contrast agent spatial distribution within plaques in CEUS image sequences. It consists of three algorithms including cardiac cycle retrieval and sub-sequence selection, temporal mean image segmentation, and texture feature extraction. The first algorithm automatically retrieves and selects cardiac cycles from CEUS frames without electrocardiogram gating. The second is composed of three steps, i.e., temporal averaging, interactive plaque delineation, and automatic neovascularization segmentation. The third extracts eight texture features from the grayscale temporal mean images and the binary segmented images. The capability of the quantitative features in discriminating between qualitative grades is examined via the t-tests, analysis of variance (ANOVA), Fisher criterion of inter-intra class variance ratio and logistic regression classification with leave-one-out cross-validation. Experimental results on 33 carotid plaques demonstrated that the optimal feature, namely the combined area ratio, exhibited significant difference among three qualitative grades (P<0.001, ANOVA). When distinguishing between low-grade and high-grade plaques, the features improved the area under the receiver operating characteristic curve, sensitivity and specificity by 36.4%, 16.7%, and 11.1%, respectively, contrasted with a classic feature, the traditional area ratio. These results demonstrate the usefulness and advantage of the proposed method in quantifying the spatial distribution of contrast agents in CEUS.