Construcción de un estimador de glucosa no invasivo portátil

Abstract

This paper presents the construction of a portable non-invasive glucose estimator, which is performed as a need for diabetics, since it allows them to obtain reference levels without the need to attend a hospital and with the advantage of being taken at any moment of the day. We used the analysis of the three usable light phenomena, transmission, reflection and absorption, in the near infrared range. A neural network model was proposed that receives data such as habits and preliminary measures that are related to the suffering of diabetes. A meter was obtained that communicates wirelessly with the signal acquisition device and a web server where the neural network is run using Python. It was tested with 65 volunteer data, yielding an estimate with an average square error of 14.8 and a correlation of 0.96 with respect to the real measure.