【摘要】 目的:进一步探讨声像图纹理分析的方法评价大鼠肝纤维化模型的价值。方法:采集90只正常或肝纤维化大鼠的肝脏标准化声像图,计算灰度共生矩阵(gray level co-occurrence matrix,GLCM)的9个声像图纹理特征参数并与病理诊断比较。结果:9个参数与病理肝纤维化S0～S4分期均具有显著相关性(P<0.05)。利用其建立判别分析模型按S0～S4分期的正确率分别为83.3%、84.2%、40%、50%和88.5%,66.7%的大鼠能够被准确分期;按S0无纤维化、S1轻度纤维化、S2和S3中重度纤维化及S4早期肝硬化分组的正确率分别为83.3%、68.4%、71.4%及88.2%,75.8%的大鼠能够被准确分组。结论:基于GLCM的声像图纹理分析能够较有效的评价大鼠肝纤维化程度。更多还原

【Abstract】 Purpose:To further investigate the diagnostic value of ultrasonic texture features in rats with liver fibrosis.Methods:Ultrasonography was performed in 90 SD rats with or without liver fibrosis.The 9 textural parameters of gray level co-occurrence matrix(GLCM) were calculated from sonograms.Liver fibrosis was staged from S0 to S4 by histopathology.Results:The 9 parameters with the severity of liver fibrosis manifested significantly linear correlations(P<0.05).According to the discriminant analysis models established on them,correct classified percentage were 83.3%,84.2%,40%,50% and 88.5% respectively when staged from S0 to S4,66.7% of cross-validated rats were accurately staged,and when categorized S0 as no fibrosis,S1 as mild fibrosis,S2 and S3 as moderate to severe fibrosis and S4 as early cirrhosis group,correct classified percentage were 83.3%,68.4%,71.4% and 88.2% respectively,75.8% of cross-validated rats were exactly grouped.Conclusion:As a convenient and accurate method,sonograms texture analysis based on GLCM can be used to evaluate rats liver fibrosis effectively.更多还原