iGLU 5.0: A Novel, Non-invasive and Intelligent HbA1c Measurement Device using Glucose values and Physiological Parameter for Smart Healthcare

Abstract

The laboratory test process of HbA1c measurement is a time-consuming and invasive method. The HbA1c parameter is the most important feature to predict the level of diabetes. Although invasive methods are irritating in the case of frequent measurements. Moreover, HbA1c measurement is only possible at the diagnostic centre, followed by medical protocol. Hence, it is still challenging to measure the HbA1c frequently at remote locations, where diagnostic centres are not easily available. Therefore, an intelligent and new non-invasive HbA1c measurement system, iGLU 5.0, is proposed for instant diagnosis of the HbA1c value without prior measurement setup. The proposed measurement device is based on optical spectroscopy for the collection of glucose values in different formats. The glucose values have been collected in fasting, postprandial, and random formats. The glucose value has also been collected using the oral glucose tolerance test (OGTT), along with the average blood pressure value, correspondingly. These four formats of glucose values, along with blood pressure, were used to predict the estimated average glucose (eAG) using an optimized prediction model. Further, the predicted average glucose is converted into an HbA1c value using a standard formula. The eAG prediction models have been trained and validated using 2000 samples of healthy, prediabetic and diabetic people to analyze the optimized prediction model. 94% and 96% accuracy have been examined during training and cross-validation of optimized DNN model, respectively. A 0.3 mean absolute difference has been identified from predicted HbA1c values using the proposed DNN model. The novel non-invasive HbA1c prediction system is useful for instant diagnosis without irritation for smart healthcare.