Enhanced approach of machine Learning models to detect Hypohyperglycemia using a closed Loop control system based on time Series data

Abstract

Patients with type 1 diabetes mellitus (DM1) typically monitor newlinetheir blood glucose levels with finger sticks multiple times daily and newlineutilize this data to subjectively predict their future glycemia to choose an newlineappropriate approach for keeping the blood glucose pressure under newlinecontrol, including insulin dosages and other factors. Patients with type 1 newlinediabetes depend entirely on insulin injections and other external newlinetreatments to retain their blood sugar levels within a normal range. A newlinepatient with diabetes who takes the naive method to manage their newlinecondition must monitor their blood glucose levels multiple times on day newlineand night with a finger-stick test. Optimizing insulin treatment to avoid newlinehypoglycemia and hyperglycemia is a critical problem in diabetes care. newlineThe present Artificial Pancreas system extracts information from the newlinecontinuous glucose monitoring (CGM) sensor to measure blood glucose newlineconcentration levels; the control algorithm incorporated in the controller newlineautomatically adjusts the insulin infusion rate of the quotautomated insulin newlineadministration system.quot Dealing with the distortion induced by food newlineintake and physical activity is one of the most significant difficulties of newlineclosed-loop glucose regulation. It employs CGM measures; however, newlinehypoglycaemia or hyperglycaemia is associated with many other newlineindicators such as carbohydrate level, energy expenditure depending on newlinephysical activity, etc. newlineFirstly, we proposed a closed-loop control system to screen the newlineblood glucose level of diabetic patients that integrates continuous newlineglucose, insulin on board, carbohydrate, and physiological variable to newlinemanage glucose levels to treat hyperglycaemia and avoid hypoglycaemia newlineby suggesting an appropriate insulin infusion rate. Using a 4-variate time newlineseries data such as glucose level, insulin dose, physical activities, and newlinefood consumption, we estimate time-varying coefficients of glucose newlineix newlinelevels in patients with type 1 diabetes using an adaptive Kalman filter newline(AKF).To test the efficacy of the suggested strategy