In the non-invasive sensing of blood glucose by near-infrared diffuse reflectance spectroscopy, the spectrum is highly susceptible to the unstable and complicated background variations from the human body and the environment. In in vitro analyses, background variations are usually corrected by the spectrum of a standard reference sample that has similar optical properties to the analyte of interest. However, it is hard to find a standard sample for the in vivo measurement. Therefore, the floating reference measurement method is proposed to enable relative measurements in vivo, where the spectra under some special source-detector distance, defined as the floating reference position, are insensitive to the changes in glucose concentration due to the absorption effect and scattering effect. Because the diffuse reflectance signals at the floating reference positions only reflect the information on background variations during the measurement, they can be used as the internal reference. In this paper, the theoretical basis of the floating reference positions in a semi-infinite turbid medium was discussed based on the steady-state diffusion equation and its analytical solutions in a semi-infinite turbid medium (under the extrapolated boundary conditions). Then, Monte-Carlo (MC) simulations and in vitro experiments based on a custom-built continuous-moving spatially resolving double-fiber NIR measurement system, configured with two types of light source, a super luminescent diode (SLD) and a super-continuum laser, were carried out to verify the existence of the floating reference position in 5%, 10% and 20% Intralipid solutions. The results showed that the simulation values of the floating reference positions are close to the theoretical results, with a maximum deviation of approximately 0.3mm in 1100–1320nm. Great differences can be observed in 1340–1400nm because the optical properties of Intralipid in this region don not satisfy the conditions of the steady-state diffusion equation. For the in vitro experiments, floating reference positions exist in 1220nm and 1320nm under two types of light source, and the results are quite close. However, the reference positions obtained from experiments are further from the light source compared with those obtained in the MC simulation. For the turbid media and the wavelengths investigated, the difference is up to 1mm. This study is important for the design of optical fibers to be applied in the floating reference measurement.
Read full abstract