Understanding the morphology of cellular organelles is one of the prerequisites to understanding their function. Shape can help explain why do certain biological processes work in a specific manner. In this work, we present statistical shape modelling of mitochondria and Golgi apparatus. For each organelle we measured the distances to the characteristic points on their surface and performed statistical analysis. We obtained approximated mitochondria shape by extracting skeleton and measuring distances from skeleton points to the surface. For Golgi apparatus, we extracted each cisternae and measured distances from their centers to the edges in each characteristic direction. Obtained results are used to generate new shapes, new distances are sampled from the statistical distribution of the measured distances which can be further refined by adjusting input parameters. New shapes are validated by comparing the distances from objects in test set. Additionally, we attempt to generate a visually similar shape of a sample not included in the statistical analysis. Derived mitochondria shapes present closely approximate real shapes, while derived Golgi apparatus shapes a little less so, possibly due to multiple assumptions and simplifications made during statistical analysis.
|
