Estimating the blood glucose of serum optical spectrum using support vector regression

Blood glucose monitoring is very important for individuals with diabetes due to its rate determining role in medication strength adjustment and observation of possible life-threatening hypoglycemia. Of the many sensor modalities tried, the combination of electrical and optical measurement is among the most promising for continuous measurements. The traditional single optical method of acquiring data was simple. The complexity of blood components, and the influence of external factors, affected accuracy of the blood glucose. We proposed an accurate computational intelligent approach using support vector regression models to estimate blood glucose concentrations of serum samples by a multi-sensor system, based on near-infrared (NIR) absorption spectrum, rotating spectrum, Raman spectrum. The results are shown that prediction data meet the clinical accuracy.