ABSTRACT We report the preliminary results obtained for estimation of glucose concentration (GC) in premature newborns, the observed glucose values were in the range of 30 to 100 mg/dL. Near infrared spectroscopy (NIR) and multivariate analysis techniques like Partial Least Square (PLS) and Principal Component Analysis (PCA) were used. A prediction model for glucose concentration was built and calibrated from a non diabetic adult volunteer and applied to neonate’s NIR spectra. In addition, we compare the glucose prediction values obtained of NIR spectra using PLS, and the values obtained with two commercial glucometer (regularly used for a rapid monitoring in premature neonates) against the clinical reference values (gold standard). Our results show that is feasible to predict the glucose concentration in neonates applying a prediction model built with an adult data. We have obtained a correlation coefficient of prediction (r 2) of 0.76 and a RMSEP of 9.5 mg/dL for the predicted results using NIR and PLS. These results are better than the obtained ones for the two commercial glucometers which showed an r 2 of 0.63 and 0.65, with RMSEP values of 12.3 and 12.1 mg/dL respectively.
Buy this Article
|