目的通过建立直径≥1 cm胆囊隆起样病变超声诊断的Logistic回归模型，筛选有助于鉴别此类病变良、恶性的超声特征。资料与方法回顾性分析165例经病理证实的直径≥1 cm胆囊隆起样病变的超声特征，包括病灶数目、大小、形态及基底宽窄，是否合并胆囊结石，胆囊壁是否连续，彩色多普勒血流显像是否检出血流信号等，通过多因素回归分析建立二分类Logistic回归模型，评价Logistic回归模型预报此类病变良、恶性的效能。结果经过二分类Logistic回归分析，病变形态、基底宽窄、彩色多普勒血流显像是否检出血流信号3个特征变量进入Logistic回归模型，是鉴别诊断胆囊隆起样病变良、恶性的敏感指标。二分类Logistic回归模型预报直径≥1 cm胆囊隆起样病变良、恶性的准确度、敏感度、特异度分别为97.0%、93.8%、97.3%，ROC曲线下面积为0.979。结论二分类Logistic回归分析能够筛选出对鉴别诊断直径≥1 cm胆囊隆起样病变良、恶性有意义的超声特征，病变形态、基底宽窄及血流信号对鉴别诊断病变的良、恶性有重要价值。%Purpose To establish Logistic regression model of gallbladder lesions of≥1 cm in diameter diagnosed by ultrasound, and to filter benign and malignant sonographic features. Materials and Methods The sonographic features were retrospectively analyzed in 165 patients with gallbladde apophysis lesions of≥1 cm in diameter which confirmed by pathology, including the number of lesions, size, shape and basal width, gallstones, continuous gallbladder wall continuous, blood flow signals detected by color Doppler flow imaging. Logistic regression model with bipartition was established by multivariate Logistic regression analysis, and the efficiency of Logistic regression model was evaluated to predict benign or malignant of these lesions. Results Three characteristic variables, including lesion morphology, basal width and flow signals, were took into the Logistic regression model by binary Logistic regression analysis, which was the sensitive indicators can differentiate the benign or malignant gallbladder lesions. The accuracy, sensitivity and specificity of this model were 97.0%, 93.8%and 97.3%for predicting the benign or malignant gallbladder apophysis lesions≥1 cm in diameter, respectively. Area under ROC was 0.979. Conclusion Binary Logistic regression analysis can filter the sonographic features which can differentiate the benign or malignant gallbladder apophysis lesions≥1 cm in diameter, and lesion morphology, basal width and flow signals are of important differential diagnosis value of benign lesions or malignant lesions.
展开▼
