Statistical Engineering Division Seminar

Dr. Antonio Possolo
Statistical Engineering Division, ITL, NIST
Dr. XiaLan Li
Information Access Division, ITL, NIST
Statistical Engineering Division Seminar
Thursday, March 27, 2008, 1:30-2:30 PM
Building 227, Room A202

Abstract

The ability to collect large numbers of microphotographs of cell populations, especially those that are grown in culture, for the purpose of characterizing them quantitatively, poses multiple challenges in areas of digital image analysis and statistical modeling.

Such characterization depends on the cellular features that are stained for imaging, and on the modality of microscopy that is employed; and it includes not only the distribution of cell size, shape, and orientation relative to the substrate, but also the pattern of the cells' spatial arrangement relative to one another.

Since identifying and delineating cells is a typical first step in this endeavor, we begin by reviewing some of the approaches that we believe are most promising for this purpose. In particular, we emphasize the power of contextual approaches to this task.

We suggest that deformable templates are useful to characterize cell shape, not only owing to their intrinsic flexibility, but also because they involve parsimonious, parametric representations that lend themselves to calibration to empirical data using conventional statistical methods. In addition, deformable templates can serve as a useful building block in models for cell motion.

To describe the patterns of interaction between cells that are captured in still microphotographs, we use the theory of stochastic spatial point processes, and illustrate the use of corresponding descriptive statistical tools, as well as parametric models that can be fitted to data, either for single populations, or for multiple populations that co-exist in the same culture, and that summarize those interactions.