A MACHINE LEARNING APPROACH TO QUANTIFY DISSOLUTION KINETICS OF POROUS MEDIA

Microspheres are popular drug products comprises of drug particles embedded in a matrix of wax and pore former. Microstructural attributes of these beads affect the overall dissolution release kinetics of the product. Due to the complex geometry and high computational cost associated with pore-scale simulations, the impact of microstructural attributes on the drug release rate is yet to be well studied. In this paper, we propose a machine learning framework to examine the drug release rate by estimating the temporal profile of the effective diffusion coefficient of the dissolved drug through the pores. By incorporating a statistical description of the pore structure via the Minkowski functionals, our model can also provide probabilistic distribution of the effective property at a given time. Leveraging such efficient numerical framework, we conduct sensitivity analysis and rank the geometric parameters according to their impacts on the drug release rate.