Artificial neural network for the voltamperometric quantification of diclofenac in presence of other nonsteroidal anti-inflammatory drugs and some commercial excipients

This work presents a methodology for the quantification of diclofenac using differential pulse voltammetry and an artificial neural network model. A carbon paste-multiwalled carbon nanotubes electrode is used as working electrode for obtaining the analytical data. Diclofenac was electrochemically characterized and the voltammetric parameters were optimized by means of the modified Simplex method, obtaining a LOD of 0.74 umol L-1 and a LOQ of 3.49 umol L-1. With the optimized parameters, an artificial neural network model was used for the quantification of diclofenac in presence of paracetamol and naproxen as REDOX pharmaceutical interferences along with other chemicals used as part of the pharmaceutical excipients. Voltammograms obtained for different concentration combinations of these drugs were compressed with a Wavelet Discrete Transform. The architecture of the mathematical model is based on a multi-layer perceptron network and a Bayesian training algorithm. With the trained model, an R-2 of 0.96 is obtained for the test data when quantifying the drugs, allowing an extrapolation of the calibration curve for the diclofenac quantification. Five pharmaceutical samples were tested, yielding a R-2 of 0.98 and a 98.1% recovery percentage for diclofenac.