Ultrasonic Tissue Characterization by Texture Analysis Using Normal and Pathologic Livers and a Reference Phantom: Relationship Between Depth Range, Scanning Plane, Probe Frequency, and Magnification Rate in the Ultrasound Image
王 怡, 伊東 紘一, 谷口 信行, 川井 夫規子, 中村 みちる, 横田 京子
Yi WANG, Kouichi lTOH, Nobuyuki TANIGUCHI, Fukiko KAWAI, Michiru NAKAMURA, Kyoko YOKOTA
Department of Clinical Pathology, Jichi Medical School
Co-occurrence matrix, Liver diseases, Texture analysis, Tissue characterization, Ultrasonography
The co-occurrence matrix method was used for the texture analysis of variously diseased livers and a phantom. Twenty cases of normal, fatty, and cirrhotic livers, each confirmed by clinical findings, laboratory data, surgery, or biopsy, were included in the study. Because of such confounding factors as magnification rate, probe frequency, and depth of the region of interest (ROI) in the ultrasonographic image, no significant differences were noted between the three types of liver. To correct for the effects by these factors, 60 parameters of texture analysis were measured using a homogeneous internal echo phantom. Changing magnification rate, frequency, and scanning plane, we measured the distribution of each parameter at different depths and found that if other conditions were unchanged, the parameter distributions were almost the same at different depths. Distribution may be used as a correction coefficient to correct for error in measurement resulting from these factors.