主成分Logistic回归在早期鉴别卵巢囊肿研究中的应用及其软件开发

Logistic Regression Based on Principal Component Analysis in Ovarian Cyst's Earlier Period Differential Diagnosis and its Software Development

目的:应用主成分L og istic回归分析方法对卵巢囊肿进行早期鉴别诊断,并对此实施软件开发以便于门诊辅助诊断,进而提高疾病鉴别诊断的准确性和效率。方法:首先对L og istic回归模型进行共线性诊断,然后筛选出对鉴别卵巢囊肿有统计学意义的检查指标并建立L og istic回归模型,利用该模型对卵巢囊肿进行早期鉴别诊断。程序开发选用delph i 7.0软件。结果:孕次、产次、流产次数间存在着中等程度的共线关系;除孕次、形态、内壁结构外,其余11项指标均纳入模型;回顾性判别符合率为87.86%,前瞻性差别符合率为85.14%,判别效果较好。结论:主成分L og istic回归实现了卵巢囊肿的早期鉴别诊断,开发的软件可以用于门诊辅助诊断。

Objective: To adopt logistic regression model based on principal component analysis to discriminate and diagnose ovarian cyst in earlier period,for which we developed the software in order to make clinical ausiliary diagnosis easy and to raise the accuracy and effectiveness of differential dianosis of disease.Methods: firstly,making collinearity diagnosis on logistic regression model,then screening the examination indexes with statistical significance for ovarian cyst differential diagnosis and establishin the logistic regression model,adopting the model to discriminate and diagnose ovarian cyst in earlier period.delphi 7.0 was selected.Results: there was a moderate degree of collinearity among gravidity,parity and abortion,the eleven indexes enter the model expect for the parity,morph and endospore texture.The retrospective discrimination coincidence rate was 87.86%,the prospective discrimination coincidence rate was 85.14%,the effect of discrimination was better.Conclusion: ovarian cyst's earlier period differential diagnosis was realized by logistic regression model based on principal component analysis.the software developed can be used in clinic as the auxiliary diagnosis.