Abstract

Movement of biological cells is crucial to a wide range of physiological processes. This has motivated the development of 3-D flexible models of living biological cells in slime molds undergoing shape changes during locomotion, imaged by using optical sectioning fluorescence microscopy. While living cells are strongly structured, they are not rigid and therefore exhibit high variability. A fundamental task is construction of models that incorporate both variability and structure in a mathematically precise way. For this we use the global shape models of Grenander. In this approach an ensemble of biological shapes are represented in terms of their typical structure via the construction of templates, and their variabilités are represented by the definition of probabilistic transformations applied to the templates. The transformations form groups and are applied so that while a rich family of shapes may be generated from a single template, the global properties of the templates are maintained. We take a Bayesian approach, with the optics of the sectioning microscope modeled as in Preza et al. and with the CCD camera modeled as a non-homogeneous Poisson counting process. The algorithms are implemented on a massively parallel SIMD architecture.

© 1992 Optical Society of America