颈部淋巴结超声特征的计算机提取及相关性评价

目的：评价计算机所提取的颈部淋巴结超声特征与医生的经验判断的相关性，以预测能否实现颈部淋巴结超声图像计算机辅助诊断。材料和方法：利用彩色多普勒超声观察43例患者的106个淋巴结的超声特征，包括大小、形态、边界、纵横比、淋巴门、皮髓质分布及比例、淋巴结内部回声、血流分布等。将超声医生判断分级的结果与计算机提取的淋巴结超声特征的参数定量结果作相关分析。结果：相关性较好的参数为NRL过零率（r=0．304，P=0．002）、凸包深度（r=0．32，P=0．001）、长／短径比率（r=-0．318，P=0．001）、血流像素点数（r=0．702，P〈0．001）、投影轮廓点分布的熵（r=0．719，P〈0．001）。结论：通过对颈部淋巴结超声图像提取量化特征参数，初步实现了颈部淋巴结良、恶性疾病的计算机辅助诊断。

Correlative Evaluation and Computer Aided Features Extraction of Cervical Lymph Nodes Ultrasonic Image

Purpose: To make correlative evaluation of cervical lymph nodes features between diagnosis results from doctors and computer aided extraction system. And make a forecast estimation of this computer aided diagnosis system.Materials and Methods: Ultrasonic features (size, margin, border; echogenic homogeneity, courter, shape, echogenic level, medulla ratio, medulla distribution, vascular density and vascular pattern) of 106 cervical lymph nodes from 43 patients were evaluated with color Doppler ultrasound equipment and computer aided extraction system respectively. Correlations between the quantitative feature parameter and the doctors' diagnosis were calculated to assess the effectiveness of these parameters.Results: Good correlations between the results from computerized scheme and doctors were found in features of size, nodal border, shape, echogenic level, medulla ratio, medulla distribution, vascular density and vascular pattern. Conclusion: The computerized feature parameters may be used as assisted indication for differentiating benign and malignant cervical lymph nodes.