Statistical image analysis for a confocal microscopy two-dimensional section of cartilage growth
Images are the source of information in many areas of scientific enquiry. A common objective in these applications is the reconstruction of the true scene from a degraded image. When objects in the image can be described parametrically, reconstruction can proceed by fitting a high level image model. We consider the analysis of confocal fluorescence microscope images of cells in an area of cartilage growth. Biological questions that are posed by the experimenters concern the nature of the cells in the image and changes in their properties with time. Our model of the imaging process is based on a detailed analysis of the data. We treat the true scene as a realization of a marked point process, incorporating this as the high level prior model in a Bayesian analysis. Inference is by simulation using reversible jump versions of Markov chain Monte Carlo algorithms which can handle the varying dimension of the image description arising from an unknown number of cells, each with its own parameters.
Keywords: Confocal fluorescence microscope; Image analysis; Marked point process; Markov chain Monte Carlo methods; Object recognition; Reversible jump Markov chain Monte Carlo method; Stochastic simulation
Document Type: Research Article
Publication date: 2004-01-01