A unified geometric model for 3D confocal image analysis in cytology

In this paper, we use partial differential equation based analysis as a methodology for computer-aided cytology. We wish to accurately extract and classify the shapes of nuclei from noisy confocal microscopy images. This is a prerequisite to an accurate quantitative intranuclear (genotypic and phenotypic) and internuclear (tissue structure) analysis of cancerous and pre-cancerous specimens. We study the use of a geometric-driven scheme for improving the results obtained by a nuclear segmentation method, based on automatic segmentation, followed by object reconstruction and interactive classification. We build a chain of methods that includes an edge-preserving image smoothing mechanism, an automatic (albeit non-regularized) segmentation method, a geometry-driven scheme to regularize the shapes and improve edge fidelity, and an interactive method to split shape clusters and reclassify them.