KIP publications

We present a reagent-free approach for long-term continuous glucose monitoring (cgm) of liquid samples using midinfrared absorption spectroscopy. This method could constitute an alternative to enzymatic glucose sensors in order to manage the widespread disease of Diabetes. In order to acquire spectra of the liquid specimen, we use a spectrally tunable external-cavity (EC-) quantum cascade laser (QCL) as radiation source in combination with a fiber-based in vitro sensor setup. Hereby we achieve a glucose sensitivity in pure glucose solutions of 3 mg/dL (RMSEP). Furthermore, the spectral tunability of the EC-QCL enables us to discriminate glucose from other molecules. We exemplify this by detecting glucose among other saccharides with an accuracy of 8 mg/dL (within other monosaccharides, RMSEVC) and 14 mg/dL (within other mono- and disaccharides, RMSECV). Moreover, we demonstrate a characterization of the significance of each wavenumber for an accurate prediction of glucose among other saccharides using an evolutionary algorithm. We show, that by picking 10 distinct wavenumbers we can achieve comparable accuracies to the use of a complete spectrum.

We present a reagent-free approach for long-term continuous glucose monitoring (cgm) of liquid samples using midinfrared absorption spectroscopy. This method could constitute an alternative to enzymatic glucose sensors in order to manage the widespread disease of Diabetes. In order to acquire spectra of the liquid specimen, we use a spectrally tunable external-cavity (EC-) quantum cascade laser (QCL) as radiation source in combination with a fiber-based in vitro sensor setup. Hereby we achieve a glucose sensitivity in pure glucose solutions of 3 mg/dL (RMSEP). Furthermore, the spectral tunability of the EC-QCL enables us to discriminate glucose from other molecules. We exemplify this by detecting glucose among other saccharides with an accuracy of 8 mg/dL (within other monosaccharides, RMSEVC) and 14 mg/dL (within other mono- and disaccharides, RMSECV). Moreover, we demonstrate a characterization of the significance of each wavenumber for an accurate prediction of glucose among other saccharides using an evolutionary algorithm. We show, that by picking 10 distinct wavenumbers we can achieve comparable accuracies to the use of a complete spectrum.