Creating a Low Cost Non-Invasive Blood Glucose Monitoring System with an Artificial Neural Network and Microcontroller

Overview

Diabetes mellitus is the ninth leading cause of death worldwide, affecting millions worldwide and likely more due to inefficient testing, and is growing worse every year. The main way to treat diabetes is with blood samples and insulin injections; however, this method is problematic, as current methods of detecting when insulin injections are required can be inaccurate, painful, expensive, wasteful, and intrusive. These methods are expensive and typically reserved for richer countries and individuals. The research goal of this study was to develop a low-cost, non- invasive, and accurate glucometer. A ESP32-CAM microcontroller and laser diode were utilized. The camera and laser diode use laser spectroscopy to determine the concentration of an individual’s blood glucose. Deep learning with logistic regression was used to calculate the estimated glucose level of an individual due to its effectiveness in estimating values based on an image. The effects of image count and image resolution on the accuracy of the neural network were investigated. An average accuracy of 90.06% (SD 6.64, n=10) was achieved using multiple high-quality images. The embedded system involves a non-invasive, pain- free, lightweight, and low-cost alternative to the current methods of glucose detection and insulin injection. The system presented could provide access to essential healthcare to millions worldwide at a low cost and with a non-invasive method. The general methodology could also be expanded to creating wearable biomonitoring technology for healthcare and other fields such as the military.

Competition history

• JSHS 2023 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.