Advancements toward noninvasive glucose sensing using near-infrared Fourier transform spectroscopy

Abstract

<p>Advancements toward noninvasive glucose sensing using Fourier transform near-infrared spectroscopy is the focus of this dissertation. The results of the Diabetes Control and Complications Trial (DCCT) conducted in 1993 underscore the importance of tight glycemic control through the use of frequent blood glucose monitoring. Since then, substantial research efforts have been made to advance glucose monitoring technologies. Noninvasive glucose sensing has long been a target of many researchers as a painless alternative to the traditional finger-stick method of glucose monitoring and could provide a means for continuous glucose monitoring without the inconvenience of changing sensors and the complications associated with the foreign body response toward implants.</p> <p>The research described in this dissertation explores novel methods to enhance our fundamental understanding of the complexity of noninvasive spectroscopy of human skin. The primary theme of this work centers on the analysis of background spectral variance related to changes in the skin tissue over a time period of several hours. The magnitude and impact of background spectral variance are examined and characterized for sets of noninvasive skin spectra collected from human volunteers. Systematic changes in the background variance are shown to negatively impact accuracy of concentration predictions from calibration models generated with the partial least-squares (PLS) and net analyte signal (NAS) algorithms. Such variations prevent accurate glucose prediction outside of the calibration region. Future work to improve the quality of glucose concentration predictions from noninvasive near infrared spectroscopy requires characterization, and ultimately control, of the background spectral variance in order to realize robust multivariate calibration models that function over time post-calibration.</p>