丁子 清

Kiyoshi CHOJI

北海道大学医学部放射線医学

Department of Radiology, Hokkaido University School of Medicine

キーワード : Ultrasound diagnosis, Chronic hepatitis, Liver cirrhosis, Multivariate statistical analysis

This paper presents a new method of diagnosis of chronic hepatitis and cirrhosis based upon the statistical analysis of findings observed in ultrasonographic images. The author applied a method of multivariate discriminant analysis using the categorical data for six significant findings reported in Part 1 of this paper. Initially, analysis was performed on the basis of all findings: Hepatic Vein Narrowing (HVNARRW), Hepatic Vein Irregularity (HVIRREG), Echotexture of the Liver (ECHOTEXTURE), Surface Echo Irregularity (SURFACE), Liver Blunting (BLUNTING), and Spleen Size (SPLEEN). Further analyses were conducted by selecting various combinations of findings, in order to determine their feasibility for clinical diagnosis. It was revealed that ECHOTEXTURE is an indispensable finding. Optimum diagnostic results are obtained when combinations of three or more findings, including ECHOTEXTURE and HVNARRW, are used as input to the statistical analysis. Based on these results, it was determined that a practical method of diagnosis is capable of being implemented.