A geometric model for 3-D confocal image analysis

In this paper, we use partial-differential-equation-based filtering as a pr eprocessing add post processing strategy for computer-aided cytology, We wi sh to accurately extract and classify. the shapes of nuclei from confocal m icroscopy images, which is a prerequisite to an accurate quantitative intra nuclear (genotypic and phenotypic) and internuclear (tissue structure) anal ysis of tissue and cultured specimens. First, we study the use of a geometr y-driven edge-preserving image smoothing mechanism before nuclear segmentat ion. We show how this biter outperforms other widely-used filters in that i t provides higher edge fidelity. Then we apply the same filter,,vith a diff erent initial condition, to smooth nuclear surfaces and obtain sub-pixel ac curacy. Finally we use another instance of the geometrical filter to correc t for misinterpretations of the nuclear surface by the segmentation algorit hm. Our prefiltering and post filtering nicely complements our initial segm entation strategy, in that it provides substantial and measurable improveme nt in the definition of the nuclear surfaces.