A segmentation algorithm for measuring blood glucose in hand-held devices

Abstract

This thesis deals with the estimation of blood glucose values from biomedical images. This estimation is related to the chemical reaction between glucose and a chemical reactant. The main goal is to segment the blood sample area from several sets of test measurements. An algorithm based on histograms thresholding and its combination with watershed segmentation is proposed and applied to these images. Furthermore, the kernel density estimator is used, as an alternative to the histogram, to estimate the probability density function of the images. After the identification of the optimal thresholding methods, a general procedure is proposed. This procedure includes image preprocessing, blood sample detection, histogram thresholding, binary masks comparison and a relative remission estimation. Then, using a convergence criterion the convergence value of the chemical reaction is found. Finally, the results are evaluated with respect to both accuracy and computation time