On the role of perfusion index for estimating blood glucose levels with ultrasound-assisted and conventional finger photoplethysmography in the near-infrared wavelength range

Abstract

Background and objectivesThe perfusion index (PI) has recently attracted considerable interest and started to be investigated as a cost-effective diagnostic tool due to its non-invasive character. However, its potential as a painless marker of blood glucose concentration remains to be explored. This study aimed to assess the usefulness of perfusion index (PI) in estimating blood glucose using conventional and ultrasound-assisted finger photoplethysmography (PPG) in the near-infrared (NIR) wavelength range. MethodA customized module combining a reflective PPG sensor with a peak emission wavelength of 940 nm and an ultrasonic sensor with an operating frequency of 40 kHz was developed. 4-min PPG recordings (two minutes with ultrasound; two minutes without ultrasound) were acquired from 53 volunteers, including 19 diabetic patients, after conventional blood sample collection via an invasive glucometer. We computed Pearson's coefficient (r), the mean absolute error (MAE), Bland-Altman plots, and Clarke error grids to explore the relationship between PI and glucose values and evaluate the performance of the proposed approach, with and without ultrasound. ResultsWhen ultrasonic waves were applied, PI showed a higher correlation with blood glucose levels (r = −0.90, p < 0.001), MAE significantly decreased (p = 0.0051), the bias was slightly closer to zero, the LOA interval became narrower, and predicted glucose levels deviating from reference values by more than 20 % decreased in number. ConclusionsThere is an association between PI and the blood glucose level, and 40 kHz ultrasound can significantly increase the accuracy in estimating the blood glucose level through PI.