In Vivo Ultrasonic Parametric Imaging of Carotid Atherosclerotic Plaque by Videodensitometric Technique

Abstract

Extensive experimental and clinical data show that the ultrasonic image conveys infor mation on the biochemical composition of the atherosclerotic plaque, ie, the relative content of lipids (hypoechoic), fibrous tissue (hyperechoic), and calcific deposits (very echogenic with shadowing). A more dishomogeneous echo structure of the plaque is also more often associated with clinically complicated carotid plaques. To date, however, the assessment of plaque density and homogeneity by transcutaneous B-mode imaging remains subjective and qualitative. The aim of this study was to assess whether plaque echodensity and homogeneity might be established on a more objective and quantitative basis by description of the spatial distribution of echo amplitude (referred to as tissue texture) applied to digitized images, obtained with commercially available B-mode transcutaneous imaging systems. A total of 47 B-mode images derived from echotomographic studies in 10 patients were digitized off line. For each region of interest, a set of first-order (mean gray level, standard deviation, skewness, kurtosis: mathematical descriptors of the shape of the frequency distribution of gray-level histogram) and of second-order (entropy, angular moment: mathematical descriptors of the spatial distribution of gray levels within the region of interest) textural parameters were evaluated. The visual, concordant reading by two independent, experienced observers assigned the plaques on the basis of qualita tively assessed echodensity to three groups: "soft" (n = 18), "fibrotic" (n = 20), "calcific" (n = 9). Regarding spatial gray-level distribution, 46 plaques would be separated into "homogeneous" (n = 17) and "dishomogeneous" (n = 29). On digitized images, the normalized mean gray level was the most effective first-order textural parameter for distinguishing soft (24.2 ± 12.4 arbitrary units in a zero to 255 scale) from fibrotic (64.5 ±16.4) and calcific plaques (125.3 ±24.5), P<0.01 for all intergroup differences. "Homogeneous" plaques were separated from "heterogeneous" ones on the basis of entropy (5 ±1 vs 7.9 ±9.7; P<0.01), whereas the values of angular second moment overlapped (1.542E-3 + 1.334E-3 vs 5.181E-4 ±2.5615E-4, P = ns). In conclusion, quantitative texture analysis of ultrasonic images derived from tran scutaneous, high-resolution, commercially available B-scan systems is feasible in man and provides a quantitative operator-independent assessment of plaque echodensity and homogeneity.