ANN-assisted UV imaging for non-destructive dissolution prediction of HPMC matrix tablets

This study investigates the use of UV imaging combined with artificial neural networks (ANNs) to estimate the dissolution behaviour of extended-release hydroxypropyl methylcellulose (HPMC) matrix tablets. UV illumination enables the detection of HPMC, allowing polymer-specific optical information to be extracted. Caffeine was used as the model active pharmaceutical ingredient, and HPMC served as the matrix-forming polymer. Formulations were prepared with HPMC contents ranging from 5% to 35%. The ANN models were trained using colourimetric information extracted from UV images of the tablets. Among all models evaluated, the one based on the blue channel of the RGB colour space showed the best performance, achieving an average f₂ similarity factor of 80.68 on the dissolution curves of the external validation set. To evaluate the performance of the trained model on an external test set under dynamic conditions, powder blends with different HPMC contents were introduced sequentially, resulting in gradual changes in tablet composition. The trained model reflected these formulation changes in the predicted dissolution profiles, capturing the shifts in release behaviour associated with the varying HPMC levels throughout the experiment. UV imaging provided relevant input for dissolution modelling and enabled the rapid, non-destructive assessment of formulation-dependent drug release. The results therefore extend the applicability of imaging-based methods to excipient-level formulation changes.