Blood glucose prediction using non-invasive optical system based on photoplethysmography

Abstract

Several people must frequently evaluate their blood glucose since it is an important indicator of health problems mainly diabetes. Different medical systems are commercialized to measure blood glucose levels; some are invasive others are noninvasive. The main purpose of this article is to develop a non-invasive device for measuring blood glucose levels based on the detection and analysis of the photoplethysmogram signal. The developed systems include an optical sensor to detect the photoplethysmography (PPG) signal, digitalizing and acquiring boards to a computer and a software program to process and analyze the digitalized PPG signal regarding some features extracted from its waveform. These features are the systolic amplitude Sa and the b/a amplitude ratio in the second derivative PPG (SDPPG) waveform. An invasive glucometer is also used along with the Sa and b/a ratio determined from the developed system to generate a calibration model which is used to deduce blood glucose level (BGL) values. The result showed that the calibration model using the b/a ratio is more accurate for non-invasive blood level measurement then that of Sa with a difference in glucose estimation around 2 mg/dl and with the correlation coefficient (R2) of the glucose level prediction between 0.8904 and 0.9775.