Design of a mechanical-tunable filter spectrometer for noninvasive glucose measurement.

The development of an accurate and reliable noninvasive near-infrared (NIR) glucose sensor hinges on the success in addressing the sensitivity and the specificity problems associated with the weak glucose signals and the overlapping NIR spectra. Spectroscopic hardware parameters most relevant to noninvasive blood glucose measurement are discussed, which include the optical throughput, integration time, spectral range, and the spectral resolution. We propose a unique spectroscopic system using a continuously rotating interference filter, which produces a signal-to-noise ratio of the order of 10(5) and is estimated to be the minimum required for successful in vivo glucose sensing. Using a classical least-squares algorithm and a spectral range between 2180 and 2312 nm, we extracted clinically relevant glucose concentrations in multicomponent solutions containing bovine serum albumin, triacetin, lactate, and urea.