GluCam: Smartphone Based Blood Glucose Monitoring and Diabetic Sensing

Abstract

Diabetes occurs when the blood glucose level is higher than the nominal value, and for its multitude sensing, both invasive and non-invasive techniques are used. Notable drawbacks include dearth of laboratory facilities and personnel at remote locations, invasive techniques, additional equipment resulting in exorbitant costs, lack of smartphone implementation, and non-autonomous functioning. The paper addresses these problems and proposes gluCam - a novel, autonomous, non-invasive, optical-based, smart-diabetic sensing model. It automatically segments blood vessels and calculates the tortuosity measure. Using the tortuosity measure and time from meal intake, a regression polynomial is developed for predicting the blood glucose level. Diabetes is diagnosed if the tortuosity measure is less than 1.15. For effortless smartphone implementation, gluCam incorporates image processing techniques to quantify blood glucose levels. Our model reports a sensitivity of 94.28%, specificity of 82.61%, mean absolute error of 10.7%, and an overall accuracy of 91.89% (for 81 participants). The model remains unaffected by lighting conditions and independent to device platform. It thereby manifests itself as a definitive and appropriate substitution for the invasive laboratory blood glucose tests by buttressing the property of self-diagnosis of diabetes.