Artificial neural network prediction of TiO2-doped chitosan micro/nanoparticle size based on particle imaging measurements

Abstract

In this study, TiO₂-doped chitosan micro/nanoparticles were fabricated using the ionic gelation mechanism under several process parameters to exhibit the strategy of introducing particle image data for the prediction of particle size. Herein, we report on a detailed methodology for the prediction of prepared particles via artificial neural network (ANN) algorithm using the multi-layer perceptron (MLP) and radial basis function (RBF) models to select the model that demonstrates the best performance for estimation of particle size. Chitosan and TiO₂-doped chitosan micro/nanoparticles were imaged, processed, and analyzed as particle diameters in order to explore prediction models, which were developed under three different classes of prepared particles (chitosan, TiO₂-doped chitosan, and chitosan/TiO₂-doped chitosan). Models were built using particle fabrication process parameters as input with particle size as output. The established MLP model successfully predicted the particle size of all classes with the mean square error (MSE) and correlation coefficient (R) between the observed and predicted values in the range of 0.0012–0.0065 and 0.85–0.90, respectively. The best results for prediction were achieved from the RBF model for all classes of particles where MSE and R values were determined as 2.93 × 10⁻²²–4.93 × 10⁻¹¹ and 1.0, respectively. Results successfully highlighted the prediction process of particle sizes via MLP and RBF models could be relevant in the decision to produce TiO₂-doped chitosan particles and confirmed the usefulness of particle image data for simulation.

Graphical Abstract