Bipolar Disorder Recognition Research Based on Fisher Discriminant Analysis
Zhang Hao-we · Zhongguo yixue wulixue zazhi · 2015
Objective It is generally accepted that bipolar disorder and unipolar depression are two different diseases for they have different pathogenesis and clinical treatment. There is a big difficulty in early identification between bipolar disorder and unipolar depression. This paper is to discuss the application of Fishier discriminant analysis(FDA) method in early recognition of bipolar disorder, in order to achieve the purpose of distinguishing between unipolar depression and bipolar disorder at the early stage. Methods 251 cases were selected in this study, including 144 cases with recurrent depression and 107 cases with bipolar disorder. General demographic dada and clinical features were statistically analyzed between the two groups. 9 variables with significant differences were screened as the observation variables of the Fisher discriminant analysis. Part of the sample were randomly selected as the training sample, the other as the test sample. Fisher linear discriminant function model was established according to the training sample, and the back substitution estimation method was used for calculating the misjudgment rate. The established FDA model can be further used for prediction of the test sample. Results The difference of age, gender, educational level and family status between the two groups had no statistical significance. While the difference of HAMD, HAMA, YMRS, MDQ and HCL was statistically significant. The Fisher linear discriminant function model was established based on 9 observation variables, the misjudgment rate of this model for the training sample was 18.4%, the forecast accuracy for the test sample was 74.2%. Conclusion The test results of Fisher discriminant function model are in good agreement with those obtained from the actual situation, and the Fisher discriminant function can be used for early recognition of bipolar disorder.