Application of the Mixtures of Experts algorithm for signal processing in a noninvasive glucose monitoring system

The theory of Mixtures of Experts (MOE) [M. Jordan, R. Jacobs, Hierarchical mixtures of experts and the EM algorithm, Neural Computation 6 (2) (1994) 181-214; S.R. Waterhouse, D.J.C. MacKay, et al., in: D.S. Touretzky (Ed.), Bayesian methods for Mixtures of Experts, Advances in Neural Information Pr ocessing Systems, Vol. 8, MIT Press, Cambridge, MA, 1996, pp. 351-357; S.R. Waterhouse, Classification and regression using Mixtures of Experts, PhD T hesis, Cambridge University, Cambridge, 1997] was applied to the signal fro m a noninvasive glucose monitor for the purpose of converting raw signal da ta into blood glucose values. The MOE algorithm can be described as a gener alized predictive method of data analysis. This method uses a superposition of multiple Linear regressions, along with a switching algorithm, to predi ct outcomes. Any number of input/output variables are possible. The unknown coefficients in this method are determined by an optimization technique ca lled the Expectation Maximization (EM) algorithm. The noninvasive GlucoWatc h(R) biographer operation has been described [R.T. Kurnik, B. Berner, et al ., Design and simulation of a reverse iontophoretic glucose monitoring devi ce, J. Electrochem. Sec. 145 (12) (1998) 4119-4125]. Briefly, a small elect rical current results in the transport of glucose beneath the skin to a hyd rogel placed on the skin surface. Within the hydrogel, the glucose reacts w ith the enzyme glucose-oxidase to produce hydrogen peroxide. This hydrogen peroxide then diffuses to a platinum-based electrode, where it reacts to pr oduce a current. The integral of this current (charge) over the sensing tim e is the signal used to measure extracted glucose. This process is repeated , yielding up to three measurements per hour. The data used for this analys is were obtained from diabetic subjects wearing the biographer over a 15-h period. The MOE inputs consisted of elapsed time, integrated current, blood glucose value at the calibration point, and a calibrated signal. The outpu t was the value of blood glucose at each measurement. These training data w ere used to determine the unknown parameters in the MOE by the Eh? algorith m. Using a 3-h time point for calibrating the biographer, the mean absolute error (MAE) between the actual blood glucose and the blood glucose predict ed with the MOE, was 14.4%. (C) 1999 Elsevier Science S.A. All rights reser ved.