卵巢肿块性质的超声指标Fisher判别分析

Fisher discriminant analysis of ultrasonography index in diagnosing characters of ovarian neoplasms

目的建立超声指标诊断卵巢肿块性质的Fisher判别函数。方法共纳入卵巢非赘生性肿块48个、良性肿瘤137个、恶性肿瘤120个,以超声参数和多普勒血流信号为鉴别诊断变量,逐步判别分析法建立Fisher判别函数,绘制投影图和领域图,回代法和刀切法验证。结果①单因素分析显示3类肿块在体积、内部舒张期血流速度(VED)、平均血流速度(Vm)、阻力指数(RI)、搏动指数(PI)、物理性质、回声、形态、边界、腹水和血流信号分级方面差异有统计学意义;②逐步判别分析法显示肿块体积、RI、物理性质、形态和边界是鉴别诊断肿块性质的显著指标。两个非标化Fisher判别函数是:方程1=0.002体积-4.793RI+0.468物理性质+0.862形态+0.901边界-4.076;方程2=0.005体积-1.480RI+0.851物理性质-0.291形态+0.443边界+0.524;③3类卵巢肿块的二维坐标投影散点基本清晰;④回代法验证函数诊断非赘生性肿块、良性肿瘤和恶性肿瘤的敏感度分别是91.67%、88.32%和93.33%;刀切法分别是91.67%、86.13%和93.33%。结论超声肿块体积、RI、物理性质、形态和边界是鉴别卵巢肿块性质的显著指标,Fisher判别分析法可为卵巢肿块性质提供较好的鉴别诊断模型。

Objective To construct Fisher discrminant functions with index of ultrasonography. Methods A total of 48 non-neoplastic ovarian cysts,137 benign and 120 malignant ovarian tumors were enrolled in this study. Taking ultrasonographic parameters and Doppler blood flow signals as differential diagnosis variable,a diagnosis model was developed using stepwise discriminant analysis. Then a projection and territorial map were drew and the diagnostic ability of the model was verified with substitution method and jackknife. Results ①Univariate analysis indicated that ovarian cysts volume,end-diastolic blood flow velocity （VED）,mean blood flow velocity （Vm）,resistance index （RI）,pulse index （PI）,physical property,echo,shape,boundary,ascites and blood flow signal have statistical difference among the three kinds of ovarian cysts. ②Stepwise discriminant analysis showed that volume,resistance index,physical property,shape and boundary are the independent prognostic variables. The two Fisher discriminant functions were as following：Function 1=0.002volume-4.793 RI＋0.468physical property＋0.862shape＋0.901boundary-4.076,Function 2=0.005volume-1.480 RI＋0.851physical property-0.291shape＋0.443boundary＋0.524. ③The projective positions of three kinds of ovarian cysts at 2D coordinates were clear. ④The sensibility and specificity of mode for diagnosis non-neoplastic ovarian cysts,benign and malignant ovarian tumors was 91.67%,88.32% and 93.33% with substitution method,and was 91.67%,86.13% and 93.33% with jackknife method. Conclusion Cysts volume,RI,physical property,shape and boundary are the significant differential prognostic variables. Fisher discriminant analysis can provide a reliable prognostic model for ovarian cysts.