Empirical Modeling of Enteric-Coated Mini-Tablet Dissolution Performance Utilizing X-Ray Computed Tomography and Convolutional Neural Networks

Mini-tablets have been utilized as an alternative to monolithic tablets due to their ease of use for pediatric populations, dose flexibility and tailoring of drug release profiles. Similar to monolithic tablets, mini-tablets can develop film coating and internal defects during manufacturing processes that may adversely affect their dissolution performance. The use of X-ray Microcomputed tomography (XRCT) is well documented for monolithic tablets as a means of identifying internal defects, but applications to mini-tablets have not been well studied. In this study, we have developed a computational program that analyzes reconstructed XRCT images of enteric-coated mini-tablets using deep learning convolutional neural networks to automatically segment individual mini-tablets and quantify individual mini-tablet physical parameters. This algorithm was utilized to determine the correlation between physical features of mini- tablets, such as micro-crack volume and enteric coating thickness, and two-stage dissolution performance in individual mini-tablets. The correlations were then used to predict the two-stage dissolution performance of individual mini-tablets based on the physical parameters obtained from XRCT images. The analysis and results provide insight into the physical variability of mini-tablet populations that are generated during manufacturing, enabling optimization of critical tableting and coating parameters in order to achieve the target dissolution criteria.