Topographic modeling of cellular images

Modeling the three-dimensional (3D) microscopic cellular images analytically is rather a difficult task due to their random shapes and deformable characteristics. One remedy is to use the topographic structures to approximate the sample surfaces and produce the unknown molecular structures by means of deformable shape generation methods from the topographic models. Here, a training sample set of 3D images is collected for shape discrimination. Morphological watersheds are applied to isolate the cells from the surrounding background. Each detected particle is enclosed within a bounding-box for contrast independent analysis. Topographical structures are adopted to model particle classes and the classification is performed in minimum Euclidian-distance sense. Our experiments show that cell images can be identified consistently in topographic structure means.