Noninvasive blood glucose level monitoring for predicting insulin infusion rate using multivariate data

Diabetes stands as the most widely recognized acute disease globally, resulting in death when it is not treated in an appropriate manner and time. We have developed a closedloop control system that uses continuous glucose, carbohydrate, and physiological variable data to regulate glucose levels and treat hyperglycemia and hypoglycemia, as well as a hypoglycemia early warning module. Overall, the proposed models are effective at predicting a normal glycemic range from >70 to 180 mg/dl, hypoglycemic values of <70 mg/dl, and hyperglycemic value of 180 mg/dl blood sugar levels. We undertook a seven-day, day-and-night home study with 15 adults. Initially, we started with checking insulin levels after meal consumption, and later, we concentrated on how our system reacted to the physical activity of the patients. Evaluation was conducted based on performance parameters such as precision (0.87), recall (0.87), F-score (0.82), delay (26.5±3), and error size (1.14±2).