Artificial neural network approach to estimate the composition of chemically synthesized biphasic calcium phosphate powders

Abstract

Calcium phosphates ceramics are widely used as implants and scaffolds in different orthopedic and dental applications. Major chemical and physical properties of these bioceramics; such as solubility, biodegradation, and mechanical behavior; are dependent on their structures, which in turn are mostly determined by the calcium to phosphate (Ca/P) ratio in their empirical formula. In this study, the Ca/P ratio of wet chemically synthesized calcium phosphate powders was estimated using a three-layered back-propagation neural network (BPNN). Biphasic calcium phosphate (BCP) samples were synthesized via wet chemical method to prepare the training and testing data sets. The pH and Ca/P ratio of the reactants were considered as inputs and the Ca/P ratio of the products was considered as the output parameter. BPNN was then optimized by changing the number of samples for each stage and the number of hidden layer neurons. The accuracy of the optimized network was tested using additional empirical samples that had not been used in the training stage. The synthesized powders were characterized by X-ray diffraction and Fourier transform infrared spectroscopy techniques. Comparing the results of the predicted values and the experimental data indicate that the developed model has an acceptable accuracy in estimating the Ca/P ratio of the BCP powders, and that a proper correlation between the actual and calculated values exists. By predicting the Ca/P, synthesis condition can be tailored to obtain desired properties according to the biomaterial application requirements.