Presenter: Daniel Skomski

Co-Authors: Zdenek Moravek, Pankaj Aggarwal, Somya Singh, Wei-Liang Chen, Randy Crawford, Ti-Chiun Chang, Jerry Klinzing, Antong Chen, Guglielmo Iozzia, Roberto Irizarry

In drug product manufacturing of solid dosages, a critical quality risk is the propensity towards major failures such as tablet lamination or fracturing during manufacturing or use. Some relevant attributes that need to be carefully monitored include the formation of cracks and agglomerates in tablets. However, such features are often analyzed only qualitatively or semi-quantitatively and conventional analysis methods can be time-consuming, inconsistent, and subjective with substantial examiner bias. To address the need for fully quantitative and consistent defect analysis, an automated method was developed that performs defect detection in tablets from X-ray Computed Tomography (XRCT) data. The method achieves the quantitative assessment of defects in terms of defect type, number, and size of each defect per tablet, for both cracks as well as agglomerates. This automated tool elucidates and de-risks boundary conditions of manufacturing failures. We compare various analytics approaches for accuracy and precision and discuss deployment via Amazon cloud computing.