Phase 2 Accomplishments

Noninvasive Blood Glucose Analysis Using Near Infrared Absorption Spectroscopy

 

Kamal Youcef-Toumi and Vidi A. Saptari

The objective of this project is to realize a noninvasive blood glucose measurement system that would significantly aid in the treatment of diabetes. Among several other optical methods, near-infrared absorption spectroscopy has been studied to be the most promising one to date. Despite tremendous amount of research for the past several years, there has not been any breakthrough that would enable the realization of a practical noninvasive glucose monitoring system. The problems that have prohibited the success of this technique can be expressed in one form: low signal-to-noise ratio.

In this project sensitivity analysis of glucose in aqueous solutions was performed. The concentration levels studied spanned between 3.3mmol/L and 100mmol/L. In the first-overtone region (5500-6500cm-1), glucose absorption at ~5930cm-1 has been found to be the most useful. The first-overtone region was selected because of the relatively lower water absorption compared to the longer wavelength region, such as the combination band region (4200-4800cm-1). This would permit the radiation to penetrate deeper into the tissue and the blood vessels. However, glucose absorption in the combination band region is still a possible alternative and is worth further investigations due to its stronger glucose absorption intensities. Due to the weak and broad glucose absorption spectral features in the near-infrared region, the real absorbance spectra are usually masked by spectral baseline variations. To minimize the effects of low-resolution baseline variations and higher resolution noise, digital Fourier-filtering was applied to the absorbance spectra. Figure 1 shows the absorbance plots for clinically relevant glucose concentrations in 10-mm path-length aqueous solutions, after the application of a Gaussian-shaped digital filter. We then attempted to analyze the absorption intensity as a function of concentration. The intensity was taken to be the peak-to-peak height of the Fourier-filtered spectrum. Figure 2 shows the results, which implies a linear relationship, with a gradient of 1.1´10-5 per mmol/L. Careful analysis indicated that the real absorbance intensity was twice the filtered absorbance value. Hence, for every 1mmol/L change in glucose concentration, there is a 2.2´10-5 change in absorbance. To complete the sensitivity analysis, spectral noise over the relevant wavelength range had to be measured. The root-mean-square noise value was measured to be ~3.3´10-5 for the 10-mm path-length solutions between 5850cm-1 and 6000cm-1 wavenumbers. Therefore, this gives a signal-to-noise ratio of 0.67 for each 1mmol/L change in glucose concentration.

Back to List of Accomplishments