Application of artificial neural networks for Process Analytical Technology-based dissolution testing

Abstract

This work proposes the application of artificial neural networks (ANN) to non-destructively predict the in vitro dissolution of pharmaceutical tablets from Process Analytical Technology (PAT) data. An extended release tablet formulation was studied, where the dissolution was influenced by the composition of the tablets and the tableting compression force. NIR and Raman spectra of the intact tablets were measured, and the dissolution of the tablets was modeled directly from the spectral data. Partial Least Square (PLS) regression and ANN models were developed for the different spectroscopic measurements individually as well as by combining them together. ANN provided up to 3% lower root mean square error for prediction (RMSEP) than the PLS models, due to its capability of modeling non-linearity between the process parameters and dissolution curves. The ANN model using reflection NIR spectra provided the most accurate predictions with 6.5 and 63 mean f₁ and f₂ values between the computed and measured dissolution curves, respectively. Furthermore, ANN served as a straightforward data fusion method without the need for additional preprocessing steps. The method could significantly advance data processing in the PAT environment, contribute to an enhanced real-time release testing procedure and hence the increased efficacy of dissolution testing.

Keywords: Artificial neural network; Dissolution prediction; NIR spectroscopy; Process Analytical Technology; Raman spectroscopy; Real-time release testing.