Comparison of Mid and Near-Infrared Spectroscopy to Predict Creatinine, Urea and Albumin in Serum Samples as Biomarkers of Renal Function

Abstract

It is relevant to develop new technologies to enable the intensive monitoring of kidney function in a minimally invasive, rapid, and economic way. With this goal, the current work compared infrared spectra in the mid- (MIR) and near-infrared (NIR) regions to predict the biomarkers of kidney function, i.e., serum creatinine, urea, and albumin. After the evaluation of diverse spectra pre-processing methods and spectral regions, it was possible to develop, in either spectral region, good to excellent regression models to predict these three biomarkers, with determination coefficients (R2) of over 0.9 and relatively low root mean squared error (RMSE). The two techniques are complementary, since the MIR spectroscopic platform presents the advantage of enabling high-throughput analysis, as it includes the sample analysis based on a micro-plate with 96 wells, only requiring a simple dehydration step before spectra acquisition, while the NIR spectroscopic platform enables the direct analysis of the serum solutions in real time. Therefore, both platforms present complementary characteristics with a high potential to enable reagent-free and frequent monitoring of kidney function and incorporation into point-of-care diagnostic systems.