2016 Volume 36 Issue 1 Pages 33-40
We constructed estimation process for time series variation of insulin by only glucose values based on mathematical model of OGTT in previous study in order to evaluate glucose metabolism for patients without insulin data in OGTT. The parameters in the mathematical model for the target patient were defined with Gaussian noise based on other patient's parameters which were identified by glucose and insulin values. The constructed method selected the estimated insulin with similarity between estimated and real glucose level from the defined model parameters. For validation of the constructed method, estimation experiments of insulin levels were executed for OGTT data of 141 patients in Kochi Medical School hospital based on cross-validation approach. The insulin estimation results showed that the true insulin values were contained in the estimated insulin values described as distribution and that characteristics of estimated insulin distribution differed from each glucose metabolism situation. The constructed method supported a certain evaluation of glucose metabolism with estimated insulin level distribution by only glucose level based on other patient's data.