颈部淋巴结超声图像特征提取的评价研究

Evaluation Studies of Ultrasonic Image Features Extraction for Cervical Lymph Nodes

对颈部淋巴结超声图像特征的计算机提取进行研究,将所提取的外形特征、边界特征、皮髓质比例、皮髓质分布、长/短径比率、血流密度和血流分布等7类超声特征参数与医生的分级判断结果相比较,以评价各特征参数的有效性。对106个颈部淋巴结的分析结果显示,所提取的参数中归一化径向长度的过零率、凸包深度、长/短比率、血流密度和血流信号的轮廓投影分布熵等参数,具有较好的相关性。

Ultrasonic image features of cervical lymph nodes were extracted by a computerized scheme. The scheme quantified the parameters of shape, border, cortex to medulla ratio, distributing of medulla, long axis to short axis ratio, vasoalar density and vascular pattern features. Then correlations between the experts and the computerized scheme for assessing parameters in corresponding categories were computed to evaluate the validity of the parameters extracted by the computer. 106 cervical lymph nodes were studied, and the results showed that good agreement was seen in NRL（normalized radial length） zero crossing count, convex hull depth, long axis to short axis ratio, the vascular density and the entropy of the blood mapping boundary points＇ distribution.