Implementation of Non-Invasive Blood Glucose Monitoring System

Abstract

In the case of diabetes, measuring blood sugar levels is made challenging by the need to pick one's finger. Laboratory testing and one-touch glucometers are intrusive procedures that increase the risk of blood-related illnesses. In the current work, we propose a novel edge device with IoMT (Internet-of-Medical-Things) capabilities for precise, non-invasive blood glucose monitoring to solve this crucial issue. In this study, a NIR (Near-Infrared) spectroscopy method is used to identify the glucose molecule in human blood at two wavelengths (940 nm, and 1300 nm). The cutting-edge gadget known as iGLU is based on high-accuracy ML (Machine Learning) models and NIR spectroscopy. For accurate measurement, a DNN (Deep Neural Network) model and an ideal multiple polynomial regression model have been provided. An open IoT platform is used to evaluate the proposed gadget, and blood glucose levels are then saved there for endocrinologist remote monitoring. For device validation, the blood glucose measurements obtained from the invasive device and the projected blood glucose levels have been compared. The AvgE (Average Error) & MARD (Mean Absolute Relative Difference) of the predicted blood glucose concentration levels were determined to be 4.66 percent and 4.61 percent, respectively. There is a 0.81 regression coefficient. An accurate and economical solution for smart healthcare is offered by the suggested spectroscopic non-invasive gadget.