Development of a Non-Invasive Glucose Monitoring System Using Breath Acetone Analysis with AI-Based Estimation Model

Abstract

Diabetes is a growing worldwide health problem, with over 530 million people affected and expected to grow substantially in the coming decades. One of the main challenges in the treatment of diabetes is the painful and invasive nature of traditional glucose monitoring methods, which require repeated finger-prick blood tests during the day. This need often translates into poor compliance and irregular monitoring. To address this problem, we have created a non-invasive and cost-effective glucose monitoring system that utilizes breath acetone analysis combined with a smart microcontroller setup. Our system utilizes an MQ-138 gas sensor to detect acetone levels in the breath, which reflects a correlation with blood glucose levels. A PIR sensor detects the presence of the user and starts a countdown process before taking the breath sample. The information gathered from the sensors is processed using a light-weight AI-based algorithm to predict glucose levels, which are then displayed on an LCD screen together with health categorization (High, Normal, and Low). This approach eliminates the need for blood sampling, encourages frequent monitoring, and improves the comfort of diabetic patients. Due to the growing prevalence of diabetes, our approach provides a simple, painless, and portable method of monitoring glucose concentration, especially suitable for resource-constrained and home-care environments.