Machine Learning-Assisted Analysis of Electrochemical Biosensors

Machine learning (ML) is effective at handling multiparameter and nonlinear problems owing to its self-learning ability. ML is used in biosensors to predict the species or concentration of an analyte. In this work, the ML-assisted classification of electrochemical biosensor measurement data are presented to predict KCl and glucose concentrations. Experiments were carried out to obtain capacitance response for KCl concentrations of 10-100 mM using the electrochemical impedance spectroscopy technique. The amperometric method was used to obtain the current response for various glucose concentrations ranging from 0.01 to 5 mM. The multiple ML-based classifiers were used for the training and testing of impedimetric and amperometric datasets using MATLAB. The confusion matrices were obtained for different ML-classifiers and their performance was evaluated based on accuracy, precision, recall, and training time. The receiver operating characteristics were also examined to determine the efficiency of prediction. The neural network models were found to be the best-performing ML-based classifiers with the highest accuracy and precision for both impedimetric and amperometric datasets.