Reproducibility of computer-quantified carotid plaque echogenicity : Can we overcome the subjectivity?

We sought to assess the reproducibility, interobserver variability, and application to clinical studies of a new method for the quantitative assessment of carotid plaque echogenicity. Carotid plaques were scanned with the use of ultrasound, and their images were stored in a computer. They were normalized by assigning certain gray values to blood and adventitia, and the gray scale median (GSM) was used to quantify their echogenicity. The variability between storage media, between degrees of magnification, and between probes was assessed. The method was applied to 232 asymptomatic carotid plaques causing 60% to 99% stenosis in relation to the presence of ipsilateral CT-demonstrated brain infarcts. In all parts of the study the plaque GSM was measured before and after normalization to evaluate its effect. Interobserver agreement for the scanning process was assessed. The GSM mean difference before and after normalization for variability studies of storage media, degrees of magnification, and probes was -14.5 and -0.12, 2.24 and 1.68, and -8.3 and -0.7, respectively. The median GSM of plaques associated with ipsilateral nonlacunar silent CT-demonstrated brain infarcts was 14, and that of plaques that were not so associated was 30 (P:=0.003). The interobserver GSM difference was -0.05 (95% CI, -1.7 to 1.6). Our method decreases the variability between storage media and between probes but not the variability between degrees of magnification. It separates echomorphologically the carotid plaques associated with silent nonlacunar CT-demonstrated brain infarcts from plaques that are not so associated.