SECM Visualization of Spatial Variability of Enzyme‐Polymer Spots. Part 2: Complex Interference Elimination by Means of Selection of Highest Sensitivity Sensor Substructures and Artificial Neural Networks

Abstract

Polymer spots with entrapped glucose oxidase were fabricated on glass surfaces and the localized enzymatic response was subsequently visualized using scanning electrochemical microscopy (SECM) in the generator–collector mode. SECM images were obtained under simultaneous variation of the concentration of glucose (0–6 mM) and ascorbic acid (0–200 μM), or, in a second set of experiments, of glucose (0–2 mM) and 2‐deoxy‐D(+)‐glucose (0–4 mM). Aiming at the quantification of the mixture components discretization of the response surfaces of the overall enzyme/polymer spot into numerous spatially defined microsensor substructures was performed. Sensitivity of sensor substructures to measured analytes was calculated and patterns of variability in the data were analyzed before and after elimination of interferences using principal component analysis. Using artificial neural networks which were fed with the data provided by the sensor substructures showing highest sensitivity for glucose, glucose concentration could be calculated in solutions containing unknown amounts of ascorbic acid with a good accuracy (RMSE 0.17 mM). Using, as an input data set, measurements provided by sensing substructures showing highest sensitivity for ascorbic acid in combination with the response of the sensors showing highest dependence on the glucose concentration, the error of the ascorbic acid concentration calculation in solution containing the unknown amount of glucose was 10 μM. Similarly, prediction of the glucose concentration in the presence of 2‐deoxy‐D(+)‐glucose was possible with a RMSE of 0.1 mM while the error of the calculation of 2‐deoxy‐D(+)‐glucose concentrations in the presence of unknown concentrations of glucose was 0.36 mM.