Estimation of Blood Glucose from PPG Signal Using Convolutional Neural Network

The purpose of this paper is to develop a system that can measure the blood glucose level in a non-invasive way. Since blood glucose is an important indicator of health issues, frequent blood glucose measurement is required by a great number of people. Non-invasive methods are getting popular for such measurements since painful collection of blood is not required. In non-invasive methods, blood glucose level is recorded by exploiting the relationship of blood glucose level with various biological signals and parameters. This paper addresses the relationship of glucose level with PPG (photoplethysmoram) signal which can provide information regarding the volume of blood flowing in the test region through exposure of that particular region to infrared light of certain wavelengths. A dataset has been created by collecting the PPG signal of 30 people along with their blood glucose level, measured in invasive way both before and 2 hours after breakfast. A CNN model has been trained with this dataset to exploit the relationship of glucose level with PPG signal and thus the glucose level of blood have been measured. Results obtained from our system have been compared with corresponding glucose level,recorded in invasive way.