能量多普勒超声联合Logistic回归模型及受试者工作特征曲线对早孕期流产的诊断价值

Diagnosis of abortion with power Doppler ultrasound combining Logistic regression model and receiver operating characteristics curve during the early pregnancy

目的评价经阴道能量多普勒超声（TV—PDU）联合Logistic回归模型及受试者工作特征（ROC）曲线对不明位置妊娠患者进行鉴别分析，预测宫内早孕期流产（IUPA）的发生及其严重程度。方法应用TV—PDU检查子宫内膜（建模组），同时测定其血清β-HCG、孕酮，运用多因素h—gistic回归法对各研究变量进行分析，并建立回归模型和ROC曲线，应用ROC曲线分析确定各超声参数及临床指标预测判断IUPA的最佳阈值和评价各指标对IUPA诊断的准确性。另外选取60例患者（测试组）对该回归模型的真实性和准确性进行验证。结果多变量模型1（M1）选择内膜下血流阻力指数（RI）、最高流速（PSV）这两个参数，而内膜回声类型、血清孕酮、停经时间则进入多变量模型2（M2）。M1、M2的ROC曲线下面积分别为0．887、0．976，约登指数和阳性似然比分别为90．8％、87．5％和12．213、8．000。M2特异度和阳性预测值均低于M1，M1的诊断准确率较高，但M1、M2对IUPA的诊断价值均较单变量明显提高。M1、M2对测试组进行回代，模型稳定性较好。结论TV—PDU能简单、无创和有效地预测早孕期IUPA的可能性，内膜下血流动力学参数与IUPA联系密切。

Objective To estimate the diagnostic value of the patients of intrauterine pregnancy abortion （IUPA） of pregnancies of unknown location （PUL） with transvaginal power Doppler ultrasound （TV-PDU） combining Logistic regression model and receiver operating characteristics （ROC） curve. Methods Endometria of PUL were detected by TV-PDU during the early pregnancy. Serum β -HCG and progesterone were analyzed meanwhile. All variables were researched by multiple factors Logistic regression. Logistic regression equations were established. ROC curve was constructed and the cut-off value of endometrial sonographic findings and clinical data to diagnose intrauterine pregnancy abortion were studied. The accuracy of every univariate and muhivariable were evaluated and the validity and stability of the Logistic regression equation were tested. Results The RI value and PSV were included in the multivariate model 1 （M1）, while the serum progesterone, type of endometrial resonance and gestation age were included in the multivariate model 2 （M2）.The area under the curve of M1, M2 were 0.887, 0.976, respectively. Youden index and positive predictive value were 90.8%, 87.5% and 12.213, 8.000, respectively. Specificity and positive predictive value of M2 was lower than M1 which had a higher accuracy. Diagnostic value were increased obviously by the use of M1 and M2 comparing with univariate and the stability was all right. Conclusions Possibility of IUPA could be predicted by TV-PDU simply, atraumaticly and efficiently. A close relation is found between the hemodynamic parameter of endometria and the IUPA.