Search

The objective of the present work was to develop an artificial neural network (ANN) model to accurately predict the dissolution profile of immediate release tablets based on non-destructive spectral data. Six different tablet formulations with varying API (caffeine) and disintegrant (potato starch) concentrations were prepared. The near-infrared (NIR) and Raman spectra of each tablet were collected in both reflection and transmission modes, then principal component analysis (PCA) was conducted. The training of the ANN was performed at each hidden neuron number from 1 to 10 in order to determine the optimal number of neurons in the hidden layer. The best results were obtained when a small number of neurons (1-3) was used. In the case of all four spectroscopic methods, the average similarity values (f2) of the optimized ANN models were above 59 for the validation tablets, indicating that the predicted dissolution profiles were similar to the measured dissolution curves. The optimized model based on reflection Raman spectra exhibited the best predictive ability. The results demonstrated the potential of ANN models in the implementation of the real-time release testing of tablet dissolution. [ABSTRACT FROM AUTHOR]