Quantitative processed images acquired by histogram-SNR imaging used to evaluate parenchymal heterogeneity in the liver.

Abstract

To evaluate a new system for displaying processed images of liver parenchyma based on quantitative estimation of heterogeneity by texture analysis. We measured the signal to noise ratio, one of the first-order statistics in the histogram of enveloped amplitude of radio-frequency backscattered echoes, using a 3.75-MHz transducer with texture analysis in conjunction with a new method in which the small ROI (region of interest) is segmented into multiple layers to minimize the influence of tissue attenuation and beam diffraction. In our computerized system, gray-display and color-display images, two types of processed images, were produced from the visual intensity of each small ROI, which was based on its signal to noise ratio. We studied 10 cases of normal liver, 10 cases of fatty liver, and 10 cases of cirrhotic liver. The processed images obtained from these livers were reviewed to observe their features and to compare their usefulness in estimating the heterogeneity of the liver parenchyma with that of conventional B-mode images. Gray-display images of cirrhotic livers appeared much blacker than the images produced from other disorders, and color-display images of cirrhotic liver appeared much bluer or greener than the others. Rate of correct diagnosis from B-mode images was 68.3 ±6.8%; from gray-display images, 85.8±7.4%; and from color-display images, 91.7±8.2%. Rate of correct assessment from B-mode images and gray-display images was significantly correlated (p=0.0015), as was rate of correct assessment from the B-mode images and the color-display images (p=0.0060). The processed images obtained using this computerized system contributed to the correct and objective interpretation of the heterogeneity of the liver parenchyma.