A Noninvasive Glucose Estimation based on Near Infrared Spectroscopy and Pulse-Echo Ultrasound

Authors

Nandi, S., Xhan, Y., Xia, J., Singh, T., and Mastrandrea, L.

Source

RGC and ICBEM 2018 Conference, May 23-May 15, Aachen, Germany

Abstract

In this paper, combining two modes of non-invasive glucose concentration sensing to improve accuracy of prediction has been investigated. In-vitro experiments were done to establish that this method is not only possible but could also be potentially state of the art. Two different modelling techniques (Linear Regression and Gaussian Process modelling) are explored and variance weighted data fusion is used to make final predictions. Both methods have shown promise although a Monte Carlo based simulation suggested the Gaussian Process works better on average. This work provides the foundation to go further and perform in-vivo experiments and address the issues associated with that.