Premium
Thread/paper‐ and paper‐based microfluidic devices for glucose assays employing artificial neural networks
Author(s) -
Lee Wilson,
Gonzalez Ariana,
Arguelles Paolo,
Guevara Ricardo,
GonzalezGuerrero Maria Jose,
Gomez Frank A.
Publication year - 2018
Publication title -
electrophoresis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.666
H-Index - 158
eISSN - 1522-2683
pISSN - 0173-0835
DOI - 10.1002/elps.201800059
Subject(s) - calibration curve , microfluidics , analyte , glucose oxidase , chemistry , chromatography , calibration , analytical chemistry (journal) , biological system , materials science , nanotechnology , biosensor , detection limit , mathematics , biology , statistics
This paper describes the fabrication of and data collection from two microfluidic devices: a microfluidic thread/paper based analytical device (μTPAD) and 3D microfluidic paper‐based analytical device (μPAD). Flowing solutions of glucose oxidase (GOx), horseradish peroxidase (HRP), and potassium iodide (KI), through each device, on contact with glucose, generated a calibration curve for each platform. The resultant yellow‐brown color from the reaction indicates oxidation of iodide to iodine. The devices were dried, scanned, and analyzed yielding a correlation between yellow intensity and glucose concentration. A similar procedure, using an unknown concentration of glucose in artificial urine, is conducted and compared to the calibration curve to obtain the unknown value. Studies to quantify glucose in artificial urine showed good correlation between the theoretical and actual concentrations, as percent differences were ≤13.0%. An ANN was trained on the four‐channel CMYK color data from 54 μTPAD and 160 μPAD analysis sites and Pearson correlation coefficients of R = 0.96491 and 0.9739, respectively, were obtained. The ANN was able to correctly classify 94.4% (51 of 54 samples) and 91.2% (146 of 160 samples) of the μTPAD and μPAD analysis sites, respectively. The development of this technology combined with ANN should further facilitate the use of these platforms for colorimetric analysis of other analytes.