论文部分内容阅读
在光学相干层析术(OCT)无创血糖监测过程中,预测模型的建立容易受异常点的干扰。采用广义极大似然估计(M估计)建立的预测模型能够有效地通过权函数降低异常点在模型中的权重。通过人体血糖钳夹临床实验和口服葡萄糖耐量测试实验,利用M估计和最小二乘估计法(OLS估计)两种方法建立了血糖预测模型,采用交互验证法对两种模型的均方根误差(RMSE)进行了比较。对比结果表明,M估计能有效地降低血糖预测模型的RMSE值。此外,利用克拉克误差表格分析法对两个模型的预测结果进行评估,评估结果表明采用M估计建立的血糖预测模型的准确性和稳定性高于OLS估计,因此M估计更适合临床上的OCT无创血糖监测应用。
In the OCT noninvasive blood glucose monitoring, the establishment of the predictive model is easily disturbed by abnormal points. The prediction model established by generalized maximum likelihood estimation (M-estimation) can effectively reduce the weights of abnormal points in the model through the weight function. The blood glucose prediction model was established by using human blood glucose clamp clinical trial and oral glucose tolerance test using M method and least squares method (OLS estimation). The root mean square error RMSE) were compared. The comparison results show that M estimation can effectively reduce the RMSE value of the blood glucose prediction model. In addition, the prediction results of the two models were evaluated by Clarke error table analysis. The evaluation results show that the accuracy and stability of the blood glucose prediction model established by M estimation is higher than that of OLS estimation. Therefore, M estimation is more suitable for clinical noninvasive OCT Blood glucose monitoring applications.