Real-Time Glucose Level Interpretation Using a Fuzzy Logic Framework for Diabetes Management

Abstract

Successful control of diabetes necessitates continuous observation of blood glucose levels and timely intervention to prevent acute complications. Traditional threshold-based systems often fail to capture subtle glucose fluctuations, particularly in real time. This paper presents a fuzzy logic-based system for dynamically assessing diabetes status and determining insulin doses using real-time glucose data from wearable or handheld sensors. Using expert-defined linguistic variables and fuzzy membership functions, the model categorizes glucose levels into clinically meaningful states, such as hypoglycemia, normoglycemia, and hyperglycemia, with graded severity. The fuzzy inference engine generates personalized alerts and dose recommendations based on American Diabetes Association (ADA) guidelines, ensuring medical relevance. The system was implemented using Python and tested across a wide glucose range (40–310 mg/dL). Simulation results showed that the model accurately recommended 0 units at low glucose levels (50–65 mg/dL), small doses at borderline values, and aggressive dosing at critical levels, with smooth transitions between categories. Compared to traditional PID control, the fuzzy logic model offered safer, more conservative dose adjustments and reduced risk of overcorrection. Designed for integration into mobile health platforms and intelligent agents like Furhat, this model represents a major step forward in delivering autonomous, interpretable, and patient-centric diabetes care.