Fisher discriminant analysis of ultrasonography index in diagnosing characters of ovarian neoplasms / 中国医学影像技术

ABSTRACT

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 (V_(ED)), mean blood flow velocity (V_m), 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.