DEVELOPED NON-INVASIVE AND CONTINUOUS BLOOD GLUCOSE MONITORING DEVICE

• Main Author: Zakki Irfani, Muhammad
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/80946
• Institution: Institut Teknologi Bandung

Diabetes mellitus is a metabolic disease with a prevalence of 10.7 million people in Indonesia. Daily blood glucose monitoring systems can improve a person's quality of life. Traditional blood glucose measurement methods require a blood sample, making them painful and posing a risk of infection if done repeatedly. Blood glucose monitoring devices are effective tools in controlling and managing diabetes. One highly sought-after method due to its potential, accuracy, ease to use, and adaptability in blood glucose measurement is using Near-Infrared Spectroscopy (NIRS) method. One parameter related to blood glucose is the Photoplethysmography (PPG) signal. The PPG signal is obtained due to changes in blood volume and is captured using optical sensors at specific wavelengths of light. The highest blood glucose absorbance wavelength is 1600 nm. Features obtained from the PPG signal include PPGDC, PPGAC, and A. This study aims to develop a non-invasive and continuous blood glucose monitoring device. The estimation of blood glucose levels in this study uses LED light sources with a wavelength of 1600 nm to obtain PPG signals. The principle used is the NIRS method with light reflectance on the wrist. The developed device test begins with hardware and software testing, the collection of PPG signal data, usability and satisfaction testing, and full-day testing. Hardware and software testing are aimed to ensure compliance with device specifications. The collection of PPG signal data aims to create a model for estimating blood glucose levels. Usability and satisfaction data collection determine user satisfaction and device usability. Full-day testing aims to test the developed device's wearability. Additional features such as height, weight, age, gender, and subject condition are used to reduce the average absolute error value. Modeling results of all data showed that eight features play a role in reducing the mean error absolute value, which is worth 16.7889 mg/dL. Full-day testing was also conducted to test the device's performance. The average absolute error estimation of blood glucose levels for a full day is 10.542 mg/dL with 100% of distribution points in Zone A and B of the error grid analysis graph. In addition, iv the questionnaire results showed a satisfaction score of 3.45 out of 5 for overall satisfaction with the developed device.