A Nonlinear Blind Identification Approach to Modeling of Diabetic Patients

In the last decades, mathematical models have become of great importance in the context of diabetes treatment planning. Several modeling approaches based on first principles or input-output techniques have been proposed. However, a relevant open problem common to all these approaches is that they are not able to recover or to systematically account for the various unmeasured signals that affect a diabetic patient (e.g., food, physical activity, and emotions). A novel blind identification approach is introduced in this brief, allowing us to model type 1 diabetic patients and to effectively recover the unmeasured input signals. The approach is applied to an experimental study regarding identification and prediction of the blood glucose concentration in five type 1 diabetic patients.