Ultrasonographic evaluation using the internal echo in normal and cirrhotic livers: Comparison of accuracy of gray-scale and binary black-and-white images and their intraobserver reproducibility and interobserver agreement.

Abstract

We attempt to determine the possibility of classifying normal and cirrhotic livers by their internal echo texture alone, without using such basic ultrasonographic information as shape and surface character of the liver. We also assessed intraobserver reproducibility and interobserver agreement obtained using this classification to ascertain the diagnostic usefulness of the method. In this evaluation, we used both regional magnified B-mode images and binary black-and-white images, both derived from conventional B-mode images obtained from 10 patients with normal livers and 10 with cirrhotic livers. These 20 echograms were randomly divided into two groups and evaluated independently on two occasions by 12 observers who used the unaided eye and took only internal echo texture into consideration. Accuracy in distinguishing between normal and cirrhotic livers ranged from 41.7 percent to 100 percent. The intraobserver correlation coefficient r1 between evaluations of the regional magnified B-mode images was 0.63, while that for the binary black-and-white images was 0.80. Agreement between the decisions of the 12 observers in the first and second evaluations of the binary black-and-white images showedk values between 0 and 1.0 for binary black-and-white images and between -0.32 and 0.29 for regional magnified B-mode images. Subjective evaluation of normal and cirrhotic livers based on internal echo texture alone was possible: intraobserver correlation was good; and unexpectedly, agreement with black-and-white images was greater than that with the regional magnified B-mode images.