Neural Networks for Estimation of Hematocrit Density from Transduced Current Curve Patterns

The hematocrit is an important factor for clinical decision marking and the most highly influencing factor for measurement of glucose values in the whole blood by handheld devices. This paper presents the use of neural network for hematocrit estimation from the transduced anodic current curves produced by glucose-oxidase reaction in electrochemical biosensors which is used in glucose measurements. The neural network used in this paper is a single hidden-layer feedforward neural network (SLFN) trained with the derived output values collected from accurately measured values by a hospital analysis system. This method can obtain an acceptable result that can be used to reduce the dependency of hematocrit in the further steps for the measurement of glucose values in the whole blood.