Evaluation of Intelligent System to the Control of Diabetes

Diabetes, as the 4th death cause, has become a vital issue in the 21th century. However, blood sugar levels of most diabetic patients are not well controlled. For a patient with chronic disease, treatment efficiency could be influenced by the disease, remedy, and mental condition, in addition to his/her physiological status. Therefore, there is a great deal of difficulty in establishing a guideline to decide the reasonable dosage for a particular patient. To address this difficulty, this study, via team cooperation with the Department of Endocrinology and Metabolism, Taichung Hospital, has designed an artificial intelligence (AI) system. This AI system may provide a real-time monitoring of patient´s physical condition and adjust the insulin dosage from AI system to facilitate the monitoring, caring, and management of patients. Furthermore, abnormal condition may be detected earlier and then emergency treatment may be provided in time to prevent the occurrence of any unfortunate events caused by negligence. With this system, data of 4 patients was partitioned into training data set (3/2) and test data set (1/3). Training data set was entered into ANM system for an analysis of learning stage to build a prediction model for insulin dosage. Upon the completion, the test data set was tested with this prediction model for the accuracy of dosage control by bioartificial pancreas. Results showed that ANM system may effectively predict the occurrence of problems related to insulin dosage of bioartificial pancreas, with a satisfactory accuracy.