Abstract Summary (Max 250 words)
High pressure homogenization is widely applied in the food industry to disrupt cell structures in plant based beverages. Achieving sufficient disruption to prevent grainy mouthfeel often requires energy inputs that may be unsustainably high. This limitation is linked to an incomplete fundamental understanding of cell structure breakup, partly due to previous studies overlooking the heterogeneity of plant materials and the differing susceptibility of distinct cell types to homogenization. In the present study, plant based cell suspensions were homogenized across a broad range of conditions, including variations in raw materials, pressures, and number of passages. Samples were analysed using automated microscopy, and particle images were classified through a convolutional neural network (CNN) trained to distinguish morphological categories. This approach enabled quantitative assessment of how individual cell types respond to homogenization. The classifier demonstrated high precision and reliably differentiated particle morphologies. Electron microscopy provided additional confirmation by linking morphological categories to specific cell types. Results revealed clear differences in homogenization susceptibility between cell types. When combined with an analytical framework previously used for inorganic particles, the findings indicate distinct breakup mechanisms, including rupture for aleurone cells and erosion for pericarp cells. These insights contribute to a deeper understanding of cell structure disruption and offer guidance for optimizing the design and operation of high pressure homogenization processes.