The next-generation artificial pancreas is under development with the goal to enhance tight glycemic control for people with type 1 diabetes (T1D). Such technology requires the integration of a chemical sensing unit combined with an insulin infusion device controlled by an algorithm capable of autonomous operation. The work detailed in this thesis explores the potential of using near-infrared (NIR) spectroscopic sensing to serve as the chemical sensing unit.Independent measurements are demonstrated for glucose, ?-hydroxybutyrate and urea in ternary aqueous solutions using NIR spectroscopy coupled with multivariate analysis. Sensitivity and selectivity for each analyte are determined by their net analyte signal (NAS). Principal component analysis (PCA) is used to characterize the instrumental and environmental variations measured during data collection across six days. The standard errors of prediction (SEPs) resulting from the PCA-NAS calibration models are 170, 90 and 120 μM for glucose, ?-hydroxybutyrate and urea, respectively. Calibration models based on partial lease squares are shown to achieve even lower SEPs, and the selectivity of these calibration models is verified. Prediction accuracies of both calibration models are sufficient for monitoring physiological levels of glucose and urea in interstitial fluid and for the clinical detection of diabetic ketoacidosis.Noninvasive glucose measurements in people with T1D are demonstrated via PCA-NAS analysis of the NIR spectra collected on the back of the hand. Glucose levels in the subcutaneous interstitial fluid (ISF), rather than in samples of capillary blood, are shown to be better suited as reference concentrations for NIR spectroscopic sensing. Variations in non-glucose information embedded in these noninvasive NIR spectra are investigated by Hotelling's T2 test and rank annihilation, providing a basis for rejecting spectral outliers. Outlier detection and rejection by rank annihilation is shown to enhance prediction accuracy of PCA-NAS calibration models.
展开▼
