四邻接点算法分析正常肝与脂肪肝超声纹理的研究

Clinical experiment of ultrasonic texture analysis between normal liver and fatty liver by four neighborhood pixels algorithm

目的 通过临床实验检验自行设计的四邻接点算法在区分正常肝与脂肪肝超声纹理中的价值。方法 选择正常肝、临床确诊的弥漫性脂肪肝 ,用 3.5和 5 .0MHz两种探头观察肋间肝切面声像图 ,用四邻接点算法对比分析其纹理变化 ,并与直方图方法比较。结果 在两种探头条件下 ,四邻接点算法得到的相对不均匀度 (HSR)在正常肝组和脂肪肝组间都有显著性差异 ( P <0 .0 1 ) ,且差异比直方图算法大。结论 四邻接点算法能较直方图算法更准确地反映肝实质的纹理变化 ,是一个较可靠的定量指标。

Objective To evaluate the value of four neighborhood pixels algorithm designed by our lab,in the distinguishing fatty liver from normal liver.Methods Normal liver,fatty liver were chosen,and the intercostal cutting surfaces of liver were selected by 3.5 and 5.0 MHz probes respectively.By four neighborhood pixels algorithm,the changes of their texture were analyzed by contrast each other and the results were compared with those of histogram method.Results Under the both probes,the relative heterogeneity values (HSR) of normal liver and fatty liver parenchyma,which was calculated by four neighborhood pixels algorithm,were significantly different (P< 0.01 ).The difference was bigger than that of histogram.Conclusions Four neighborhood pixels algorithm is more accurate than histogram in reflecting the ultrasonic texture change of hepatic parenchyma,and it is quite a reliable quantitative parameter.