Evaluation of different segmentation methods of X-ray micro computed tomography images

Abstract

In silico tools for the prediction of disintegration and/or dissolution of tablets can be validated using adequate images of real pharmaceutical formulations. X-ray micro-computed tomography images of 12 different tablet batches prepared from binary mixtures of API and excipient were used. The goal of this work was to compare different segmentation methods to improve the results and processing time of an evaluation of pre-processing methods. The open source software ImageJ was utilised for the image processing. Different threshold algorithms were applied as well as different cluster numbers for the k-means clustering. The pathways were analysed regarding their desirability which was calculated from the recovery rates and their ratios. It was possible to identify suitable pathways for each single batch as well as for combinations of several batches. The recovery rates of the best pathways were always approximately 100%. It was possible to confirm the correctness of the image processing by visual perception. The image processing could be improved and sped up.