Noninvasive Blood Glucose Sensing Using Near Infra-Red Spectroscopy and Artificial Neural Networks Based on Inverse Delayed Function Model of Neuron

• DOI: 10.1007/s10916-014-0166-2
• Publication Date: 2014-12-10T08:08:08Z
• Authors: Swathi Ramasahayam; Sri Haindavi Kopp...; Lavanya Arora; Shubhajit Roy Cho...
• Keywords: Blood Glucose; Spectroscopy, Near-Infrared; 0202 electrical engineering, electronic engineering, information engineering; Humans; Neural Networks, Computer; 02 engineering and technology; Algorithms; Blood Chemical Analysis

In this paper, a non-invasive blood glucose sensing system is presented using near infra-red(NIR) spectroscopy. The signal from the NIR optodes is processed using artificial neural networks (ANN) to estimate the glucose level in blood. In order to obtain accurate values of the synaptic weights of the ANN, inverse delayed (ID) function model of neuron has been used. The ANN model has been implemented on field programmable gate array (FPGA). Error in estimating glucose levels using ANN based on ID function model of neuron implemented on FPGA, came out to be 1.02 mg/dl using 15 hidden neurons in the hidden layer as against 5.48 mg/dl using ANN based on conventional neuron model.