【摘要】 目的探讨Fisher判别分析对1型及2型糖尿病患者的分类判别价值。方法选取糖尿病患者165例,其中1型糖尿病病例14例,2型糖尿病病151例。以动态血糖监测系统监测患者的血糖数据计算出17个血糖波动特征,以此为原始数据建立数据库,利用SPSS统计软件提供的Fisher判别分析方法建立分类模型,检验模型判别1型及2型糖尿病患者效果,采用受试者工作特征曲线(ROC)对模型进行评价。结果建立的Fisher判别分析分类模型判别2型糖尿病的正确率为90. 1%,判别1型糖尿病组的正确率为57. 1%,1型及2型糖尿病合计判别正确率为87. 3%,交叉核实法检验总判别正确率为83. 0%。Fisher判别分析1型及2型糖尿病的ROC曲线下面积为0. 736,判别1型与2型糖尿病的准确性、特异性、敏感性分别为83. 7%、94. 4%、34. 8%。结论 Fisher判别分析对1型及2型糖尿病患者的分类能力良好。更多还原

【Abstract】 Objective To explore the value of Fisher’s discriminant analysis (FDA) in the classification of patients with type 1 and type 2 diabetes mellitus. Methods A total of 165 diabetic patients were selected,including 14 cases of type 1 diabetes and 151 cases of type 2 diabetes. Using the blood glucose monitoring system (CGMS) to monitor the patient’s blood glucose data,17 blood glucose fluctuation characteristics were calculated,and a database was established based on the original data. The classification model was established using the FDA method provided by SPSS statistical software,and we tested its ability in determining type 1 and type 2 diabetic patients. The receiver operating characteristic curve (ROC)was used to evaluate the model. Results By the classification model based on FDA method,the discriminant accuracy was 90. 1% in determining type 1 diabetes and 57. 1% in type 2 diabetes,the total discriminant accuracy was 87. 3%,and the total accuracy by cross-verified test was 83. 0%. The area under the ROC curve was 0. 736,and the accuracy,specificity,and sensitivity of the Fisher classification model in determining type 1 and type 2 diabetes were 83. 7%,94. 4%,and 34. 8%. Conclusion FDA has a good ability to classify patients with type 1 and type 2 diabetes.更多还原