Sonographic diagnosis of fatty liver using computer-aided quantification

摘要

We investigated the B-mode ultrasound (US) images of abdomen obtaining the normalized gray levels (NGL's) of the liver with the representative gray levels of the liver, adrenal cortex, subcutaneous fat, blood vessel of the hepatic portal vein (HPV) in the ultrasound images to quantify fatty interfiltration of liver (FIL). We also compared the NGL's of the liver with the clinical diagnosis of FIL showing good correlation between them. The representative gray levels of liver, adrenal cortex, subcutaneous fat, blood vessel of HPV, and blood were obtained by the various image processing methods leading to the two NGL's in the region of interest of the liver: one (NGL1) is the relative gray level of the liver between the representative gray levels of the adrenal cortex and subcutaneous fat. The other (NGL2) is the relative glay level of the liver between the representative gray levels of the blood vessel of HPV and the blood. With the US images of 95 patients ranging from normal to severe grade of FIL, the NGL1 and NGL2 of liver were compared with the clinical diagnosis of FIL. From the statistical distributions of the NGL's in the US images, we found that the NGL's increase monotonically as the clinical estimations of the severity of the FIL is severer. The calculated standard deviation of the NGL1 for the images is about 0.13. The calculated linear correlation coefficient between the NGL1 and the clinical estimation of the severity of the FIL is 0.79. The calculated linear correlation coefficient is 0.69 between the NGL2 and the clinical estimation of the severity of the FIL. When we use two NGL's of the liver simultaneously, the ambiguity in the diagnosis for the FIL could be fairly reduced compared to the case without the NGL's.