Design and Development of a Non-Invasive Glucose Monitoring Device Using Advanced Microcontroller

Millions of individuals all over the world are living with diabetes, a chronic condition that necessitates regular monitoring of Blood Glucose Levels (BGL). The present technique for minimal invasive BGL monitoring involves taking a small blood sample and transferring it to a glucometer. However, this procedure can be uncomfortable for the patient. This research presents a method for non-invasively monitoring BGL with the help of a powerful controller called the ARM Cortex-M4. For non-invasive applications, we use photoplethysmography (PPG) signals. Using a photodiode and an infrared LED light, PPG samples are received from a finger sensor. Signal processing and feature extraction are carried out using the Gray-Level Co-occurrence Matrix (GLCM) method. To predict the BGL, two ML models are developed: one employing a Linear Regression (LR) model and the other an Artificial Neural Network (ANN) model. The experimental findings show that the ANN model is quite effective, with an R2 score of 0.986. After careful consideration, the ARM Cortex-M4 microcontroller's processing capabilities were found to be optimal for running the ANN model. The model must be fine-tuned before deployment so it can communicate with the microcontroller and provide correct inference in real time. The finger sensor simplifies PPG sample collection, and the OLED display makes it simple to view BGL.