Mean-shift based algorithm for the measurement of blood glucose in hand-held devices

We propose an algorithm to estimate the glucose concentration from a very small blood sample in a novel hand-held measurement device. A photometric measurement system is used whereby a camera is employed to observe the reflectance behavior of the chemical reaction. The aim is to estimate and track this behavior in the region of interest. The approach is based on mode tracking via the adaptive mean-shift algorithm, using both spatial and range information. Finally, the estimated glucose concentration is associated with the obtained mode. We show that binning the images increases the mode detection accuracy. Using both synthetic and real data, we validate the proposed algorithm.