基于Logistic模型评价囊性卵巢良恶性病变多征象联合的鉴别诊断价值

The Value of a Combination of Multiple Signs in Differential Diagnosis of Benign and Malignant Cystic Ovarian Lesions:A Study Based on Logistic Regression Model Scoring System

目的探讨基于CT和MRI共有的影像学特征与血清肿瘤标志物CA125建立的Logistic回归模型鉴别囊性卵巢良恶性病变。方法回顾分析经病理证实的囊性卵巢良恶性病变患者72例共202个病灶,观察记录患者年龄、绝经情况、肿瘤标志物CA125、腹腔积液、肿瘤大小、形态、边界、厚壁、乳头或壁结节、强化程度等情况,用卡方检验、Logistic回归模型等方法进行分析。结果将单因素分析有统计学意义的指标(年龄、绝经、腹腔积液、CA125、肿瘤大小、形态、边界、伴有乳头或结节、强化程度)纳入多因素回归模型中,得出年龄>50岁(OR:13.725,95%CI:2.101~89.644),CA125升高(OR:7.180,95%CI:1.651~31.227)及伴有肿瘤中度及以上强化(OR:43.533,95%CI:10.312~183.785)是诊断恶性肿瘤的独立危险因素。基于Logistic模型所建立的评分系统≥3分时,曲线下面积(AUC)为0.894,敏感度为86.6%,特异度为92.6%,准确率为90.59%,阳性预测值为85.29%,阴性预测值为93.28%。结论基于影像学特征联合血清CA125建立的Logistic模型的评分系统能有效地鉴别囊性卵巢良恶性病变。

Objective To establish a logistic regression model based on common CT and MRI imaging features and serum tumor marker CA125 to differentiate benign from malignant non-solid ovarian lesions.Methods 202 cases of benign and malignant non-solid ovarian lesions confirmed by pathology were retrospectively analyzed.Age,menopause,CA125,ascites,tumor size,shape,border,thick wall,nipple or wall nodules,and degrees of enhancement were analyzed.Chi-square test and Logistic regression model were used.Results Age,menopause,ascites,CA125,size,shape,border,nipple or nodule,degree of enhancement were included in the multivariate regression model for further analysis.The results showed that over 50 years old(OR:13.725,95%CI:2.101~89.644),elevation of CA125(OR:7.180,95%CI:1.651~31.227),with moderate or higher enhancement(OR:13.725,95%CI:2.101~89.644),were independent risk factors for the diagnosis of malignant tumors.A scoring system based on Logistic model was compiled,and the area under the curve was 0.894,the sensitivity was 86.6%,the specificity was 92.6%,the accuracy was 90.59%,the positive predictive value was 85.29%,and the negative predictive value was 93.28%.Conclusion The scoring system based on Logistic regression model which combined with imaging features and serum CA125 can effectively identify benign and malignant cystic ovarian lesions.