Fuzzy logic: a contemporary technique in refining physicians’ prescribed total daily insulin dosage for type 2 diabetes patients

Abstract

Type 2 diabetes mellitus (T2DM) is one of the major global health challenges across the world with an elevated prevalence rate worldwide and responsible for 90-95% of total diabetes cases. The management of T2DM depends on conventional insulin therapy that does not have the ability to perfectly provide accurate glucose regulation inside the body. In a conventional insulin therapy, physicians prescribe insulin dosage for the patients by taking into account patientrelated factors (PRFs) i.e. weight, height, BMI or carbohydrate intake discretely or considering only one factor. The succession of insulin dosage depends on the close consideration of all the factors that have a positive correlation with the insulin sensitivity. Thus traditional insulin therapy, in turn, causes instances of hypoglycemia and hyperglycemia among the vast number of diabetes patients. This project examined 24 randomly selected T2DM patients admitted into two hospitals of Dhaka city through utilization of a fuzzy logic based technique to further tune the physicians’ prescribed total daily insulin dosage for alleviating the hypoglycemic and hyperglycemic incidences among these patients. Two patient-related factors (PRFs) such asaverage fasting blood glucose level (AFBGL) and average daily protein intake (ADPI) were considered as inputs for the fuzzy-logic system as these two PRFs have a positive correlation with insulin sensitivity. After considering insulin dose as an output variable, appropriate membership functions were defined by using MATLAB Fuzzy Logic Designer Toolbox. Furthermore, to establish a relationship among the membership functions, the ‘if/then rules’ are then set in the interface that also provides the fuzzy system with a decision-making facility. The last process known as defuzzification has enlightened the project by generating an output known to as predicted insulin dose (PID) as a recommendation by the fuzzy-based system for each patient. Through a quantitative comparison between the predicted insulin dose (PID) by the fuzzy system with the physicians’ prescribed insulin dose (PPD) for every patient, a numerical difference of different degrees was obtained indicating an additional or reduced administration of insulin dose by the patient so far causing the critical events (hyperglycemia & hypoglycemia) more prominent in their everyday life. The result of this experiment is further evinced by the data collected from those two hospitals where the experimented patients were admitted. Our experimental findings have shown a number of previous hyperglycemic and hypoglycemic events experienced by the patients as predicted by the fuzzy system respectively. Accordingly, the predicted insulin dose by the fuzzy system is believed to alleviate the hypoglycemic and hyperglycemic events in these patients in the future. Finally, a low mortality rate and a beneficial financial condition followed by a better quality of life for these 24 type 2 diabetes mellitus (T2DM) patients were possible by utilizing this prominent form of Artificial Intelligence used for precision dosing on insulin.