Dissolution prediction from images: method and validation of dissoLab platform

This paper demonstrates and validates dissoLab, a dissolution modeling software using microscopic imaging data. First principle dissolution models are solved for image voxels representing sample-specific particle sizes and morphologies. 2D images, such as SEM or PLM, can be utilized to predict dissolution profiles after a generative artificial intelligence method synthesizes a structurally similar 3D volume. Predictions can be performed from 3D datasets, such as X-ray micro-CT, without generative synthesis. Dissolution profiles predicted via this approach are validated against in vitro dissolution measurements and verified with theoretical models for particle samples.