Smart inexpensive quantitative urine glucose and contaminant bromide ion sensor based on metal nanoparticles with deep learning approach

Abstract

This present study reports a facile cost-effective colorimetric method for the estimation of glucose and bromide using colloidal silver triangular nanoprisms (STN). The quantitative glucose sensing is based on distinct color changes through the glucose oxidase enzymatic oxidation of STN at room temperature. Whereas, nonenzymatic visible bromide detection could be possible by only STN. A developed reference color scale could directly be used to estimate the different glucose concentration by bare eyes. Whereas, a smartphone-based analyte estimation app developed with You Only Look Once, Version 3 (YOLOv3) deep-learning algorithm to detect and estimate the glucose/bromide ion concentrations more accurately. With the increasing analyte (glucose/bromide) concentration, the localized surface plasmon resonance (LSPR) peak of STN shows gradual blue-shift in wavelength and decreased in absorbance. The experimental study was performed to estimate the urine glucose levels and shows high accuracy results when compared with standard methods. This highly sensitive glucose/bromide detection efficiency starts with a very low concentration of 0.010 mg/mL (1.0 mg/dL) and 0.005 mg/mL (0.5 mg/dL) for glucose and bromide ions respectively. This proposed non-invasive painless highly selective glucose sensing technique could be a promising point-of-care diagnostics or an environmental bromide monitoring tool for developing countries or in case of emergency uses.