A differential evolution based approach for estimating minimal model parameters from IVGTT data

Estimation of insulin sensitivity plays a crucial role in the diagnosis and clinical investigation of glucose related diseases. The Bergman minimal model provides a non-invasive approach for estimating insulin sensitivity from the glucose insulin time series data of intravenous glucose tolerance test (IVGTT). However, quite often in the traditional gradient based techniques for deriving insulin sensitivity from the minimal model, improper initialization leads to convergence problems and results in final solution, which are either incorrect or physiologically not feasible. This paper deals with a differential evolution-based approach for the determination of insulin sensitivity from the minimal model using clinical test data. Being a direct search based technique, the proposed approach enables the determination of the global solution irrespective of the initial parameter values. The fitting performance of the model with parameters estimated from the proposed approach is found to be higher than the corresponding model estimated from the widely used gradient based approach. A high correlation coefficient of 0.964 (95% confidence interval of [0.897,0.988]) is acheived between the estimated insulin sensitivity and the one obtained from the population based approach for 16 subjects. The high correlation signifies the relative similarity between the two estimated indices in representing the same physiological phenomena. The simulation results and statistical analysis reveal that the proposed technique provides a reliable estimate of insulin sensitivity with minimum prior knowledge.