Clinical feasibility of Raman spectroscopy for quantitative blood glucose measurement

Abstract

Diabetes mellitus is a leading cause of morbidity and mortality worldwide, and close monitoring of blood glucose levels is crucial for its diagnosis and management. Currently, blood glucose monitoring is done by blood withdrawal or through invasive access to the interstitial fluid. While Raman spectroscopy has been studied as a possible non-invasive optical technique for measuring blood glucose, it still faces several practical difficulties. These include tissue turbidity and autofluorescence, the lag between blood and interstitial fluid glucose concentrations, and the inherently weak intensity of aqueous glucose Raman signatures with respect to those of the interfering tissue. This thesis investigates the feasibility of using Raman spectroscopy as a non-invasive technique for blood glucose monitoring, and studies different strategies to overcome the barriers to clinical application. In particular, the study proposes a dynamic concentration correction scheme to correct for the calibration errors arising from the lag between glucose concentrations in the bloodstream and the interstitial fluid. In addition, Monte Carlo simulations were employed to study the differences in the distribution of Raman scattered photons along the depth of the tissue between backscattered and transmission mode Raman spectroscopy. Finally, a portable clinical Raman spectroscopy unit was developed utilizing a non-imaging optical element called a compound hyperbolic concentrator (CHC). The CHC coupled with a matching focusing lens efficiently collects and collimates Raman light from highly scattering tissues, while maintaining much smaller physical dimensions than a compound parabolic concentrator. Using the clinical instrument, skin Raman spectra were collected from healthy human subjects undergoing oral glucose tolerance tests, while the corresponding reference blood glucose concentrations were measured simultaneously with a conventional finger-stick glucose meter. From these datasets, linear and non-linear multivariate calibration techniques were used to relate the Raman spectral intensities to the glucose concentrations. The calibrated algorithms were then tested to demonstrate clinical accuracy as required by the Food and Drug Administration and the International Organization for Standardization. Despite the remaining challenges, the promising results obtained in this study provide important insights required in the clinical translation of Raman spectroscopy for non-invasive blood glucose monitoring.