3D SEGMENTATION OF WHOLE CELLS AND CELL NUCLEI IN TISSUE USING DYNAMIC PROGRAMMING

We present a highly robust, semi-automatic algorithm for segmenting individual, fluorescence-labeled, whole cells or cell nuclei (objects) from 3D confocal microscope images of tissue. The core of the algorithm is an extension of dynamic programming that finds the surface as a series of overlapping ribbons, where the mean intensity per unit area within each ribbon is a global maximum. Consequently, the segmented surface is not significantly affected by intermittent fluorescence labeling, diffuse borders or by spurious signals away from the borders. Minimal interaction by the user ensures correct segmentation of each object, yet segmenting 10 s to 100 s of objects per sample is feasible. The algorithm enables for the first time quantitative analysis and modeling of communication processes between cells in intact tissue