Diabetes Care in Motion: Blood Glucose Estimation Using Wearable Devices

Diabetics should monitor their blood glucose daily to prevent severe acute complications. However, the pain sensation due to finger pricking is not only an inconvenience to patients, but can also reduce compliance with diabetes management. A reliable, mobile, easy-to-use, and noninvasive glucometer can address this problem. Noninvasive optical signals, such as those used to obtain a photoplethysmogram (PPG), have recently been used to measure human physiological and vascular conditions and have been adopted to estimate glucose. In this study, we conducted a clinical trial to acquire the PPG signals using wearable devices from nine type 2 diabetic patients. The global and personalized models were built using random forest and adaboost regression models. The accuracy of the models was determined by the tenfold cross validation and leave-one-out validation approaches. The results show that it is feasible to attain 90% accurate glucose predictions. Therefore, diabetic patients can use a wearable noninvasive glucometer utilizing the PPG signals.