Modeling of subdivision of scored tablets with the application of artificial neural networks

Abstract

The subdivision of scored tablets is an important problem for the exact individual therapy of patients. The recent guidelines of the EU require verification of the equal mass of the tablet halves, but this problem has previously never been investigated in papers published on the production technological aspects. Our aim was therefore to study the effects of the physicochemical properties of the raw materials and the effects of the compression process on the breaking parameters of the tablets. Artificial neural networks (ANNs) were applied for data analysis and modeling, which are very useful optimizing systems. The abilities of four different types of ANNs to predict the parameters of the compression process and the tablets were compared. The radial basis function and multilayer perceptron ANNs furnished statistically significant better results than linear or generalized regression neural networks. These ANN models revealed that the subdivision of scored tablets is strongly influenced by the production parameters and the compositions of the powder mixtures. © 2009 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 99:905–915, 2010