AI-OPTIMIZED BIO-IMPEDANCE SENSING WITH PUFFER FISH ALGORITHM: A BREAKTHROUGH IN NON-INVASIVE GLUCOSE MONITORING

ABSTRACT The global rise in diabetes cases highlights the urgent need for effective, non-invasive blood glucose monitoring solutions. This study introduces a pioneering Bio-Impedance Sensing system, enhanced by Artificial Intelligence (AI) and optimized using the Bio-Inspired Puffer Fish Optimization (PFO) Algorithm, for real-time, non-invasive glucose estimation. Bio-impedance spectroscopy is utilized to detect glucose-induced changes in tissue electrical properties, while AI-based machine learning models analyse these signals to predict glucose levels with high accuracy. The Puffer Fish Optimization Algorithm plays a key role in feature selection and hyperparameter optimization, leading to improved model performance. A comprehensive dataset, consisting of bio-impedance signals and reference glucose levels obtained through standard invasive techniques, was used for model training and validation. The proposed system demonstrated a remarkable mean absolute error (MAE) of 3.2 mg/dL and a correlation coefficient (R) of 0.97, indicating exceptional predictive accuracy and reliability. This innovative approach presents a pain-free, cost-effective, and real-time solution for glucose monitoring, offering significant potential to enhance diabetes management. Future research will focus on further optimization of signal acquisition, refinement of AI models, and large-scale clinical trials to validate and enhance real-world applicability. Keywords: Bio-impedance sensing, artificial intelligence, Puffer Fish Optimization, non-invasive glucose monitoring, machine learning, diabetes management