More Accurate Non-invasive Blood Glucose Measurement Based on Dynamic Spectrum and PLS

Non-invasive blood glucose measurement has attracted much attention in recent years. Near-infrared (NIR) spectroscopy is currently used for frequent detection of blood glucose levels. The dynamic spectrum (DS) method, which can eliminate the individual discrepancy, is the development of ordinary NIR spectroscopy. Classic Least Squares (CLS) is usually employed to analyze the spectra. In this work, we apply partial least squares (PLS) regression instead to extract the glucose concentration information from spectra of human blood, which is a mixture of four main ingredients water, glucose, albumin and hemoglobin. All the calculating work was conducted by Lab VIEW. Comparing to CLS, the PLS algorithm was more precise in extracting the blood glucose concentration from the spectra. The DS method is able to reduce disturbance caused by individual discrepancy. And the PLS algorithm excels in data processing. If PLS is utilized in spectral analysis of DS, non-invasive measurement of blood glucose will surely be much more accurate.