Multiple Regression Analysis Based Non-Invasive Blood Glucose Level Estimation Using Photoplethysmography

This paper presents a non-invasive technique for blood glucose level estimation based on multiple regression analysis that uses the photoplethysmography (PPG) signal. Firstly, a digital bandpass filter is designed to remove the noise from the raw signal. Afterward, various parameters of the filtered signal are analyzed and the correlation between the parameters and blood glucose level is determined. Then, using multiple regression, a model has been established to estimate the blood glucose level. Finally, the predicted value obtained using multiple regression analysis was compared to the actual blood glucose value obtained from the invasive technique and to the predicted value obtained using simple linear regression analysis. It was found that percentage deviation lies within 8% in a multiple regression based model that ensures the model’s accuracy and for every subject, the percentage deviation in this model is less than the percentage deviation obtained in the single parameter-based model. As this research work uses a PPG signal that is comfortable to use and the accuracy of the multiple parameter based model is higher than the single parameter based model, it will be very effective to estimate blood glucose level easily and accurately.