Analyzing Blood Glucose Levels with Near Infra-Red Spectroscopy and Chemometric Multivariate Methods

TL;DR

This study improves blood glucose measurement accuracy using near-infrared spectroscopy combined with chemometric methods, specifically derivative preprocessing and PCR, to better distinguish glucose peaks amidst water interference.

Contribution

It introduces a derivative-based preprocessing technique that enhances glucose peak detection and reduces principal components in PCR, improving prediction accuracy over traditional methods.

Findings

PCR with derivative preprocessing yields smaller errors.

Derivative methods reduce water interference in spectra.

Linear regression based on molar absorptivity shows acceptable accuracy.

Abstract

In this work, the blood NIR absorbances are recorded using the FT-IR method. It is shown that when the absorbance curves are multiplied by the first derivative of the water absorbance spectrum as well as by the first derivative of the glucose absorbance, the peaks related to the water interferent in the blood are effectively removed from the blood absorbance spectra, allowing for better distinction of the peaks of the blood glucose. The PCR prediction using this method shows smaller errors compared to the PCR employing the net absorbances, while the number of derived principal components is smaller in the PCR method based on the derivatives than the one based on the net absorbances. Additionally, the prediction of blood glucose levels using a linear regression model based on the molar absorptivity of glucose also demonstrates acceptable accuracy.