Point-of-care, continuous assessment of multiple biomarkers using long wavelength near infrared spectroscopy for dialysis

Summary of abstract - Dialysis is a crucial treatment for chronic kidney disease (CKD). In this routine treatment quality care can be ensured by continuous, nondestructive monitoring to avoid clinical complications like sepsis. Long wavelength near-infrared (LWNIR) spectroscopy shows promise in providing continuous, point-of-care monitoring. This paper demonstrates the application of LWNIR spectroscopy for measuring multiple biomarkers—specifically lactate and urea— in a microfluidic system simulating a channel for dialysis. The spectroscopy calibration is based on liquid phantoms containing various interfering biomolecules. Technical abstract –Chronic kidney disease is treated with routine dialysis. Routine dialysis benefits from continuous monitoring of key biomarkers like lactate and urea in its effectiveness and early warning of complications like sepsis in patients. Current standard of care includes laboratory measurements using a sample. Long wavelength near-infrared (LWNIR) spectroscopy could potentially provide the optimal solution for non-destructive, continuous and point-of-care monitoring in peritoneal dialysis systems. However, LWNIR spectroscopy suffers today from poor calibration and specificity because of spectrally interfering biomolecules. In this work, we develop a novel method of calibrating machine learning algorithm with liquid biophantoms, in place of multiple patient samples to provide wide variability due of selected spectrally interfering molecules. Using this liquid biophantoms model we see two distinct sets of wavelengths that are relevant to determining biomarker concentration with and without the interferers. The model trained with the high variability liquid biophantoms is used to determine biomarker levels spiked serum in a microfluidic system simulating a channel for dialysis.