A Gaussian Markov random field based model for the porous structure of pharmaceutical film coatings
Drug release from oral tablets is determined by the porous structure of tablet coatings. Since the pore structure can be controlled by changing the manufacturing parameters, understanding how the pore structure affects transport properties is important for designing coatings with desired properties. We use a spatial statistical model to investigate how the transport properties of ethylcellulose/hydroxypropylcellulose coatings depend on the pore structure. The model is a thresholded Gaussian Markov random field, where the random field is non-stationary. We generate structures from the model that have varying pore sizes and shapes, and analyze statistically how the transport properties depend on the pore structure by using numerically simulated diffusion through the generated structures. We use a Markov Chain Monte Carlo algorithm to fit the model to confocal laser scanning microscope images of the coatings. The model is found to fit stationary parts of the images well.
Joint work with David Bolin and Holger Rootzén.